>
►
I
J. C. CATFORD
A Linguistic
Theory
of Translation
LANGUAGE
Q)
D
LANGUAGE
LEARNING
A Linguistic Theory
of Translation
An Essay in Applied Linguistics
J. C. CATFORD
Oxford University Press
Oxford University Press, Walton Street, Oxford 0x2 6dp
OXFORD LONDON GLASGOW
NEW YORK TORONTO MELBOURNE WELLINGTON
IBADAN NAIROBI DAR ES SALAAM CAPE TOWN
KUALA LUMPUR SINGAPORE JAKARTA HONG KONG TOKYO
DELHI BOMBAY CALCUTTA MADRAS KARACHI
ISBN O 19 43701S 6
© Oxford University Press, ig6y
First published 19 65
Fifth impression 1978
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means
electronic, mechanical, photocopying, recording or otherwise, without
the prior permission of Oxford University Press
This book is sold subject to the condition that it shall not, by way of
trade or otherwise, be lent, re-sold, hired out, or otherwise circulated
without the publisher's prior consent in any form of binding or
cover other than that in which it is published and without a similar
condition including this condition being imposed on the subsequent
purchaser.
PRINTED IN GREAT BRITAIN BY
HAZELL WATSON AND VINEY LTD
AYLESBURY, BUCKS
Preface
Translation is an activity of enormous importance in the mod-
ern world and it is a subject of interest not only to linguists, pro-
fessional and amateur translators and language-teachers, but also
to electronic engineers and mathematicians. Books and articles on
translation have been written by specialists in all these fields.
Writers on the subject have approached it from different points
of view — regarding translation as a literary art, or as a problem
in computer-programming, discussing the problem of ‘faithful-
ness’ of rendering, of whether words or ‘ideas’ are to be translated,
or of the routines to be set up, say, for stem and affix recognition
in machine translation.
The present volume is not primarily concerned with any of
these special problems, but rather with the analysis of what
translation is. It proposes general categories to which we can
assign our observations of particular instances of translation, and
it shows how these categories relate to one another. In short, it
sets up, though somewhat tentatively and incompletely, a theory
of translation which may be drawn upon in any discussion of
particular translation-problems.
Since translation has to do with language, the analysis and
description of translation-processes must make considerable use
of categories set up for the description of languages. It must, in
other words, draw upon a theory of language — a general linguistic
theory.
This book is based on lectures given in the School of Applied
Linguistics at Edinburgh University. It was thus originally
intended for an audience of students already fairly well-informed
about general linguistics. To make it more acceptable to the
general reader, an opening chapter has been added which dis-
cusses briefly the nature of language and the categories of general
linguistics as well as giving an outline of the analysis and descrip-
tion of English which underlies the discussion of a number of
examples. Parts of the book are somewhat technical. This is
PREFACE
inevitable in a book on a specialized topic, but it should not dismay
the general reader since the main arguments demand little or no
previous knowledge of linguistic science and the first chapter may
be used for reference when required.
Language-teachers, in particular, may find the book of interest.
The extent to which translation can be used in language-teaching
is an issue of great concern to teachers, and it is one which cannot
be fruitfully discussed without the support of some theory about
what translation is, about the nature of translation equivalence,
the difference between translation equivalence and formal corres-
pondence, the levels of language at which translations may be
performed and so on. The chief defect of the now almost univers-
ally condemned ‘Grammar-Translation Method’ was that it used
bad grammar and bad translation — translation is not a dangerous
technique in itself provided its nature is understood, and its use is
carefully controlled : and translation is in itself a valuable skill to
be imparted to students.
A number of students and colleagues contributed useful sugges-
tions when the essay was first circulated in duplicated draft form,
to all of whom I am grateful. In particular, however, I should
like to thank Dr M. A. K. Halliday, with whom I discussed many
parts of the work while it was in preparation, and Miss Leila
Dixon, who carried out the difficult task of typing the manuscript
in several stages.
J. C. Catford
Edinburgh, 1964
viii
Contents
] Genera! Linguistic Theory 1
2 Translation : Definition and General Types 20
3 Translation Equivalence 27
4 Formal Correspondence 32
5 Meaning and Total Translation 35
6 Transference 43
7 Conditions of Translation Equivalence 49
8 Phonological Translation 56
9 Graphological Translation 62
10 Transliteration 66
11 Grammatical and Lexical Translation 71
12 Translation Shifts 73
13 Language Varieties in Translation 83
14 The Limits of Translatability 93
1
General Linguistic Theory
1.0 Translation is an operation performed on languages: a pro-
cess of substituting a text in one language for a text in another.
Clearly, then, any theory of translation must draw upon a theory
of language — a general linguistic theory.
General Linguistics is, primarily, a theory about how languages
work. It provides categories, drawn from generalizations based on
observation of languages and language-events. These categories
can, in turn, be used in the description of any particular language.
The general linguistic theory made use of in this book is essentially
that developed at the University of Edinburgh, in particular by
M. A. K. Halliday 1 2 and influenced to a large extent by the work
of the late J. R. Firth. The present writer, however, takes full
responsibility for the brief and, indeed, oversimplified sketch of
linguistic theory given here, which differs from that of Halliday
chiefly in its treatment of levels (1,2).
1 . 1 Our starting-point is a consideration of how language is
related to the human social situations in which it operates. This
leads on to classification of levels of language (or of linguistic
analysis) and then to a discussion of the fundamental categories of
linguistics which can be used in the description of at least the
grammar and phonology of particular languages.
Language is a type of patterned human behaviour. It is a way,
perhaps the most important way, in which human beings interact
in social situations. Language-behaviour is externalized or mani-
fested in some kind of bodily activity on the part of a performer , and
presupposes the existence of at least one other human participant
in the situation, an addressee.' 1,
1 For a fuller account than it is possible to give here, the reader is referred
to M. A. K. Halliday, ‘Categories of the Theory of Grammar’, Word, Vol. 17, No.
3, 1961, pp. 241-92; also to Halliday, M. A. K., McIntosh, A., and Strevens, P.
D. ‘The Linguistic Sciences and Language Teaching’. Longmans, 1964.
2 Performer and addressee are ‘participant roles’. In the limiting case of a man
talking to himself — i.e. interacting linguistically with himself— both roles are
1
A LINGUISTIC THEORY OF TRANSLATION
The specific type of behaviour in which language is manifested
not only identifies the behaviour as language-behaviour but also
defines the medium which the performer is using. The performer’s
activity most commonly takes the form of either vocal movements
which generate sound-waves, or hand movements which leave a
visible trace. The first type of activity is a manifestation of lan-
guage in the spoken medium — the performer is a speaker, and his
addressee(s) is/are a hearer or hearers. The second type is a
manifestation of language in the written medium — the performer
is a writer, and his addressee(s) is/are a reader or readers. In the
next paragraph we shall, for simplicity, confine ourselves to
language in its spoken manifestation.
Language, as we said above, is patterned behaviour. It is, indeed,
the pattern which is the language. On any given occasion, the
particular vocal movements and the resultant sound-waves can
be described with a delicacy , or depth of detail, limited only by
the delicacy of the apparatus used for observation and analysis.
And the precise quality of these vocal movements and sound-
waves will be found to differ on different occasions, even when
the speaker is ‘saying the same thing’. From the linguistic point
of view, the important thing is that, on each occasion of ‘saying
the same thing’ the vocal activities of the speaker conform to the
same pattern.
The overt language-behaviour described above is causally
related to various other features of the situation in which it occurs.
There are specific objects, events, relations and so on, in the
situation, which lead the performer to produce these particular
vocal movements, and no others. The precise nature of the
situational features which are relatabie to the performer’s lin-
guistic behaviour will be found to differ on different occasions,
even when he is ‘saying the same thing’.
From the linguistic point of view, however, the important thing
again is that, in each case, the situational features which lead to
‘the same’ utterance conform to the same general pattern.
Language then is an activity which may be said to impinge on
the world at large at two ends. On the one hand, it is manifested
filled simultaneously by the same biological individual: but this is of the most
marginal relevance to linguistic theory (cf. 13.2).
2
GENERAL LINGUISTIC THEORY
in specific kinds of overt behaviour (e.g. vocal movements) : on
the other hand, it is related to specific objects, events, etc. in the
situation. Both of these— vocal movements, and actual events,
etc. — are outside of language itself. They are extralinguistic
events. They are the phonic substance in which vocal activity is
manifested, and the situation (or situation substance ) to which this
activity is related. The language itself is, however, the organiza-
tion or patterning which language-behaviour implicitly imposes
on these two kinds of substance — language is form, not substance.
1 .2 In order to account for language-events we make abstrac-
tions from these events: abstractions of various types, or at a
series of levels.
1.21 We distinguish, first, the levels of medium-substance ( phonic
substance, for the spoken medium, and graphic substance for the
written medium), and situation (or situation-substance), both of
which are, in fact, extralinguistic. The internal levels of language
are those of medium-form — phonology and graphology, arrived at
by a process of abstraction from phonic and graphic substance,
and the differently abstracted levels, which Halliday calls the
‘formal levels’ — grammar and lexis. 3
The relationship between (the units of) grammar/lexis and
situation (substance) is that of contextual meaning, or context.
r
f.
(Medium)
-A
Language
'I
phonic
substance
graphic
substance
—
phonology
u
grammar
—
graphology
;l
lexis
Situation
(substance)
» The term ‘formal levels’ for grammar and lexis has the inconvenience that
it suggests that no relatively independent form can be stated for the phono-
logical and graphological levels.
3
A LINGUISTIC THEORY OF TRANSLATION
The relationship between (the units of) phonology and phonic
substance has no generally recognized name, though ‘phonetic
meaning’ might be suggested. The relationship between grapho-
logy and graphic substance might likewise be called ‘graphetic
meaning’. Context is the interlevel relating grammar/lexis and situa-
tion, indicated by the dashed line on the right of the above
diagram.
1 .22 The levels at which we make abstractions from language-
events are thus the following:
1.221 Grammatical! lexical form
(i) Grammar: the level of linguistic form at which operate closed
systems : the characteristics ofa closed system being : ( 1 ) the number
of terms is finite; (2) each term is exclusive of the others; (3) any
change in the number of terms would change the ‘values’ (or
‘formal meanings’) of the other terms (e.g. systems of pronouns,
of deictics, of number, of case, of tense . . . etc.).
(ii) Lexis : the level of linguistic form at which operate open sets
(e.g. the open sets of items often occurring as examples or
‘exponents’ of nouns, verbs, etc.).
1.222 Medium form
(i) Phonology: the formal units into which phonic substance is
organized, and which operate, usually in combination, as the
exponents of grammatical/lexical forms.
(ii) Graphology: the formal units into which graphic substance is
organized, and which operate, usually in combination, as the
exponents of grammatical/lexical forms.
1 .223 Medium Substance
(i) Phonic substance : actual vocal sounds — the substance in which
phonology is manifested.
(ii) Graphic substance: actual visible marks — the substance in
which graphology is manifested.
Both types of medium substance have a certain patterning or
organization imposed upon them by medium-form.
1.224 Situation (or situation substance). All those features of situ-
ations, excluding medium substance, which are related or
4
GENERAL LINGUISTIC THEORY
relatable to language-behaviour. Situation substance has a certain
organization imposed upon it by grammatical/lexical form.
1.23 In addition, we must consider the interlevel of context (or
contextual meaning ) : the interlevel of statements about the distinc-
tive features of situation-substance which are relatable to parti-
cular grammatical/lexical forms. As we have said above, there is
another interlevel: the interlevel of statements about the distinctive
features of medium substance which are relatable to medium
forms.
It will be clear that context or contextual meaning is what is most
usually understood by ‘meaning’ : in our theory, this is only one
part of meaning, which also includes formal meaning which is the
way any item operates in the network of formal relations. Both
types of meaning are discussed in Chapter 5.
1 .3 The fundamental categories of linguistic theory — applicable
at least to the levels of grammar, phonology and probably
graphology — are unit, structure, class and system.
1.31 By a unit we mean a stretch of language activity which is
the carrier of a pattern of a particular kind. In English phonology,
for example, there is a unit, the tone-group, which is the carrier of
recurrent meaningful patterns of pitch. The following are exam-
ples of English tone-groups (the pitch-pattern being roughly
indicated by lines drawn over the texts).
X X s x
Yes. Yesterday. John came yesterday.
The fact that each of these tone-groups is a carrier of a
meaningful pattern is shown by the possibility of occurrence of
units of a similar type which differ only in that the pitch-pattern
which they carry is meaningfully different, thus :
Yes? Yesterday? John came yesterday?
In English grammar we have units such as sentence, clause and
group : each of these is the carrier of a particular kind of meaning-
5
A LINGUISTIC THEORY OF TRANSLATION
fui grammatical pattern. The following are examples of sentences,
each carrying the same pattern of arrangement of clauses.
Ill If you do that, II you will regret it. ///
HI When John arrived, // we had already started. ///
HI Having arrived too late, // we missed the start of the
concert. ///
And these are examples of clauses, each carrying the same
pattern of arrangement of groups :
II John / loves / Mary. 1 1
II The young man / was writing / a letter. //
II All these people who were here last night / were / friends of
mine. //
1.311 The units of grammar or of phonology operate in hier-
archies — ‘larger’ or more inclusive units being made up of ‘smaller’
or less inclusive units. They form a scale of units at different ranks.
Thus, the sentences quoted above each consist of two clauses. The
sentence is a unit of higher rank than the clause. And each clause
consists of several groups — the clause being a unit of higher rank
than the group.
1.32 The unit is the category set up to account for those
stretches of language-activity which carry recurrent meaningful
patterns. The patterns themselves still have to be accounted for
— and these are what we call structures. A structure is an arrange-
ment of elements. Thus, the elements of structure of the English
unit ‘clause’ are P (predicator), S (subject), C (complement),
A (adjunct).
The texts: /// John / loves / Mary. ///
HI The young man / was writing / a letter. ///
are two examples of English sentences, each of which consists of
a single clause. Each clause has the structure SPC. The following
clauses :
He / ran / quickly.
The young man / was writing / with a bail-point.
are examples of the structure SPA, and so on.
Among the units of English phonology we find the syllable : the
elements of syllable structure are N (nucleus or vocalic element),
K r (releasing (initial) consonantal element), K a (arresting con-
6
GENERAL LINGUISTIC THEORY
sonantal element), K 1 (‘interlude’ or inter-nuclear consonantal
element — occurring only between two Ns). Thus the syllables
represented in orthography by tea, car, now exemplify the structure
KN, those represented by cat, stop, lumps, etc. . . . KNK, and so on.
1.33 By a class we mean a grouping of members of a unit in
terms of the way in which they operate in the structure of the
unit next above in the rank scale. Structure, as we have said, is
stated in terms of ordered arrangements (in which linear sequence
often is, but need not always be, a characteristic) of elements:
thus, in English, the elements of structure of the unit clause are
S, P, C, A. The units which operate as exponents of these elements
are themselves groups. Groups, then, may be classified in terms of
the particular elements of clause structure which they expound.
Thus we have, in English, the class of Verbal Groups, which
operate at — or as exponents of — P in clause-structure; the class
of Nominal Groups which operate as exponents of S or C in clause-
structure, etc.
In English phonology, for instance, we have classes of the unit
phoneme, defined in terms of their operation in the structure of the
unit next above, the syllable. Thus the members of the unit
‘phoneme’, which operate as exponents of the element K r (con-
sonantal releasing element) in syllable structure constitute the
class ‘initial consonant’ or C 1 .
1.34 By a system we mean a finite set of alternants, among which
a choice must be made. Very often, these alternants, the terms in
a system, are the members of a class: thus the members of the
class ‘initial consonant’ mentioned above constitute a system of
phonemes pb t dkg . . . etc. which can alternate as exponents
of that particular class.
An example of a system in grammar might be the number-
system (Sing/Plur) (Sing/Dual/Plural), etc., of many languages.
Where number is a system of the Nominal group (as in English)
the terms in the system are themselves sub-groups or sub-classes
of the class.
1.4 We have referred already to rank (in 1.311) and have used
the terms exponent and delicacy. These three terms refer to three
scales which are part of the general theory of language, and of
language-description.
7
A LINGUISTIC THEORY OF TRANSLATION
1.41 The rank scale is the scale on which units are arranged in a
grammatical or phonological hierarchy. In English grammar we
set up a hierarchy of 5 units — the largest, or ‘highest’, on the rank-
scale is the sentence. The smallest, or ‘lowest’, on the rank scale
is the morpheme. Between these, in ‘descending’ order, are the
clause, the group and the word. By placing these in this order on the
scale of rank we mean that every sentence consists of one or more
than one clause, every clause of one or more than one group,
every group of one or more than one word, and every word of
one or more than one morpheme.
Thus ‘Yes!’ is a sentence consisting of one clause, consisting of
one group, consisting of one word, consisting of one morpheme.
And ‘As soon as the boys had arrived, their mother gave them
tea’, is a sentence consisting of two clauses. The first clause
consists of three groups, the second of four groups. In the first
clause the group as soon as consists of three words, the groups
the boys and had arrived of two words each. In the second clause,
the first group their mother consists of two words, the remaining
three groups of one word each . . . and so on.
1 .4 11 The normal relation between units in a grammatical
hierarchy is that stated here: namely that a unit at any rank
consists of one or more unit of the rank next below, or, conversely,
that a unit at any rank operates in the structure of the unit next above.
We must, however, make allowance for the fact that in all
languages we find ‘Chinese box’ arrangements of units, in which
a unit may sometimes operate in the structure of a unit of the
same or of lower rank. To deal with this, we make use of the
concept of rank-shift.
Thus, in English, clauses normally operate as exponents of
elements of sentence-structure. But we also find clauses operating
within groups , i.e. as exponents of elements in the structure of a
unit of the rank below the clause.
For example, in Since we couldn't meet earlier, we met after the
concert the clause we met after the concert is operating directly in the
structure of the sentence, as exponent, in fact, of a (a ‘free clause’)
in a sentence of structure floe (a ‘free clause’ preceded by a ‘bound
clause’) (see 1.721 below). But in The man we met after the concert
is my brother the clause we met after the concert is rank-shifted. It is not
8
GENERAL LINGUISTIC THEORY
operating directly in the structure of the sentence, but within a
Nominal Group. It is, in fact, operating as exponent of Q_
(qualifier) in the structure of the nominal group The man we met
after the concert. This nominal group, in turn, is exponent of S in
the clause The man we met after the concert (S) / is (P) / my brother (C).
Similarly in He met Susan at the party the adverbial group at the
party is operating directly in the structure of the clause — as
exponent of A. But in The girl at the party was Susan the group
at the party is rank-shifted. It is not operating directly in the clause,
but within a Nominal Group, as exponent of Q,.
The concept of rank (and rank scale) is an important one both
in theoretical linguistics and in many applications of linguistics,
including translation-theory.
1.42 The scale of exponence is a scale of ‘exemplification’ or of
degrees of abstraction, running from ‘highest degree of abstrac-
tion’ to ‘most specific and concrete exemplification’. Thus, in
English phonology, we may say that the class C (consonant)
represents the highest degree of abstraction at phoneme rank. In
any given instance, say of an utterance of the word tea, we may
say that the initial phoneme here is a (member of the class) C :
its exponent in this case is the particular phoneme / 1 /, and this,
in turn, has its ultimate exponent in a piece of actual phonic
substance, represented in phonetic transcription by, say, [t^ 1 ].
Exponence is related to rank in the sense that an element of
structure of a unit at one rank is expounded by — or has as its
exponent — a unit or units of the rank next below. But exponence
is a separate scale, and at any one rank we may go off sideways,
as it were, to a relatively concrete exemplification : thus we might
call the sequence of particular grammatical and lexical items
represented by ‘A linguistic theory of translation’ an exponent of
the unit ‘group’. In other words, we also use the term exponent in
talking of the relationship between the abstract units and items
of grammar and lexis and their realizations in medium form.
Thus, in English, I is the graphological exponent of the grammat-
ical item ‘1st person singular subject pronoun’, bank is the
graphological exponent of two different lexical items which we
might label X (meaning ‘money shop’) and Y (meaning ‘border
of river . . . etc.’) and so on.
9
A LINGUISTIC THEORY OF TRANSLATION
1.43 The third scale mentioned here is that of delicacy, this is
the scale of ‘depth of detail’. At a primary degree of delicacy, we
recognize, or set up, only the minimal number of units or classes,
etc., which are forced upon us by the data. Thus, if we are going
to attribute any structure at all to English nominal groups we must
set up three elements: H (head), M (modifier) and Q, (qualifier) .
Our least delicate description of English Ngp structure is thus
(M . . . n) H (Q, . . . n), which means that one element, H,
is always present, and this may be preceded and/or followed by
one or more element M or Q,. Thus we should say, at a primary
degree of delicacy, that the groups :
Old / men
These three old / men
have the
structure, MH and MMMH. By taking a further step down the
delicacy scale we recognize different classes of the element
M — namely d (deictic), o (numerative) , e (epithet), and we can
say that These three old / men has the structure d o e H, in
which d o e is a more delicate statement of structure than MMM.
1.5 Lexis. We stated in 1.221 that lexis is that part of language
which is not describable in terms of closed systems. The distinction
between- grammar and lexis is not absolute, but rather in the
nature of a dine, with very well marked poles, but some overlap
in between.
In English, for instance, most exponents of the word-class verb
are open-set lexical items : a few, such as can, may etc. are purely
grammatical items: and a few others are either lexical or
grammatical, e.g. BE which is a lexical item in ‘He is a teacher’ or
‘He has been a teacher.’ and a grammatical item in ‘He is talking’.
1.51 The categories discussed in 1.2 are not applicable to lexis.
We deal formally with lexis in terms of collocation and lexical sets.
A collocation is the ‘lexical company’ that a particular lexical
item keeps. Any particular lexical item tends to collocate most
frequently with a range of other lexical items. We refer to the
item under discussion as the node or nodal item, and the items with
which it collocates as its collocates. Thus in English, if we take
sheep and mutton as nodes we will find that each has a distinct range
10
GENERAL LINGUISTIC THEORY
of collocates : e.g. sheep collocates frequently with such lexical items
as field , flock, shear, etc., mutton collocates with such lexical items
as roast, menu, fat . . . etc. There are certainly overlaps in
collocational range— thus we may have a (whole) roast sheep and
we might have fat sheep as well as mutton fat, but on the whole they
have different collocational ranges, and this establishes the fact
that they belong to different lexical sets and are different lexical
items.
A lexical set is a group of lexical items which have similar
collocational ranges.
1.52 Collocation and lexical set are concepts which sometimes
enable us to establish the existence of two distinct lexical items,
even when both share exactly the same medium exponents. Thus
in English we have a graphological form bank — but the fact that
this enters into two distinct collocational ranges, and hence
apparently belongs to two distinct lexical sets enables us to say
that there are two distinct lexical items which happen to
share the same medium exponents, graphological bank, phono-
logical/ bar)k/. 4
1.6 We mentioned in 1.0 that our approach to the levels of
language and linguistic analysis was somewhat different from
that of Halliday, and indicated in 1.21 that this difference lay
in the fact that we set up a separate level of medium form. In other
words, instead of regarding phonology (and likewise graphology) as
an interlevel linking phonic (or graphic) substance directly with
the ‘formal levels’ of grammar and lexis, we regard the medium
as being to some extent autonomous and detachable from gram-
mar and lexis. Since this view of medium as ‘detachable’ is
important for our theory of translation, some justification and
discussion of it must be given here.
1.61 Medium form is a part of a language. Every language has
its characteristic phonology and many languages have a character-
istic graphology. In the process of analysing and describing a
language we set up, as phonological units, just those bundles of
* Following a widely accepted convention, phonological forms are normally
cited within slant-lines. Occasional use is, however, made of single and double
vertical lines, as in 1.61 below. These are used only when explicit reference is
being made to the description of English Phonology given in 1.71.
11
A LINGUISTIC THEORY OF TRANSLATION
distinctive phonic features which function contrastively in the
exponence of grammatical and lexical items of that language.
Thus we set up / p / and / b / as distinct phonemes because such
pairs as / pig / and / big /, / pak / and / bak / are exponents of
distinct lexical items: and we set up the foot or rhythmic unit as
a phonological unit because the difference in foot-division be-
tween such pairs as
|| that’s a | blackbird ||
and || That’s a | black | bird ||
is exponent of a difference in grammatical structure:
| blackbird | = compound-noun as H in Ngp structure,
| black | bird | = adjective + noun as MH in Ngp structure.
1.62 In other words, the discovery procedure for phonological
analysis must depend directly on grammatical/lexical differences.
But once the phonology has been established , by discovering what
phonic distinctions operate as exponents of grammatical/lexical
distinctions in that particular language, it can be regarded —
indeed must be regarded — as relatively autonomous or indepen-
dent. It is this autonomy of phonology which makes it possible
for two or more lexical or grammatical items to share the same
phonological exponents — e.g. the three or more distinct English
lexical items which share the one phonological exponent / pi 3 / —
partially distinguished in graphological exponence as peer and
pier. It also makes it possible for one single item to have more
than one phonological exponent, such as the English ‘indefinite
article’ which has the alternative phonological exponents / 3 / or
/ 3 n /, and the ‘nominal plural morpheme’ which has a series
of phonological exponents / s,z,iz /, / 3 n /, / internal vowel-change /
etc.
1.63 More striking evidence of the autonomy and detachability
of medium is the fact that the grammar and lexis of one language
can be expounded (though often with some losses in distinctive-
ness) in the medium of another. We are all familiar with the
Englishmen who speaks French fluently and ‘correctly’, but who
speaks it entirely through the medium of English phonology. His
12
GENERAL LINGUISTIC THEORY
grammar/lexis are purely French — but his phonology is English.
We normally attribute a certain primacy to grammar/lexis, since
iii this case we should say ‘He’s speaking French with an English
accent’ but not ‘He’s speaking English, but with French grammar
and lexis’.
1.64 Graphology, too, is in a sense detachable from the particular
language of which it is characteristic. The air traveller in India,
for example, notices on one side of his plane, the legend :
INDIAN AIRLINES
and on the other:
ffttfT Wells! 41
This Devanagari inscription, which might be transliterated
Ic^iyan eyarlains is exponent of a piece of English grammar and
lexis. It is English expounded in Devanagari (Hindi) graphology.
1.65 It is the detachability of the medium levels from the
gramma tical/lexical levels which makes phonological and grapho-
logical translation possible.
1.7 We have already drawn upon English for examples in this
chapter, and we will continue to do so throughout this book. It
seems desirable, therefore, to give here the barest outline of the
description of English phonology and grammar which we are
using. This is not the place to give a full description, even in
summary form, of English — but the indications given here will
serve to codify what has already been referred to, and will help
to elucidate most of the references to English given later.
1.71 English Phonology. In English phonology we have a hier-
archy of units at four ranks :
(i) Tone-group
(ii) Foot (or rhythmic group )
(iii) Syllable
(iv) Phoneme
The relation between these is the normal one: i.e. every Tone-
group consists of one or more Foot, every Foot of one or more
Syllable, every Syllable of one or more Phoneme. Thus || Yes ||
(with, say, falling tone) is a tone-group, consisting of one foot.
13
A LINGUISTIC THEORY OF TRANSLATION
consisting of one syllable, consisting of three phonemes. And
|| What did you | do | yesterday || is a Tone group consisting of
three feet. The first foot || What did you | and the last foot
| yesterday || each consist of three syllables : the middle foot consists
of only one. And the syllables consist of varying numbers of
phonemes.
1.711 The tone-group. The elements of tone-group structure are
T (tonic) which is always present, and P (pretonic) which may
be absent. The exponent of T is a foot, or more than one foot,
which carries one of a system of five contrastive tones', the dis-
tinctive tone starts on the first syllable (of the first foot) of the
tonic. The exponent of P, if present, is one or more foot preceding
the tonic, and carrying one of a restricted range of pretonic
intonation contours. In these examples tone-group boundaries
are marked by || , foot-boundaries by | , the initial syllable of the
tonic by underlining.
1.712 The location of the tonic is significant. It can be shifted
from one foot to another, and such shifts are changes of tonicity.
For example:
|| David was the | one who did | all the | work ||
j| David was the | one who did | all the | work ||
|| David was the | one who did | all the | work ||
14
GENERAL LINGUISTIC THEORY
1.713 The tone-group, then, is the unit which carries contrastive
intonation patterns. The contrasts are of two kinds (i) contrasts of
tone, i.e. selection of one or another out of a system of five tones
operating at the tonic: e.g.
1. I! yes II \ falling
2. II yes || / rising
3. || yes || / low-level -f rise
4. || yes || \y fall-rise
5. || yes || /\ rise-fall
and (ii) contrasts of tonicity, i.e. selection of one or another loca-
tion for the tonic.
1.714 The foot. This is the unit of stress or rhythm. The foot
is the carrier of contrastive differences in stress-distribution. The
distinctive phonic features of the foot are (i) each foot is ex-
pounded, or manifested, by a major chest pulse starting strongly
stressed, then falling off (stress-curve ) : if the foot con-
sists of more than one syllable, this means that the first syllable
is more strongly stressed than its successor (s), (ii) each foot with-
in one and the same tone-group tends to have approximately the
same duration.
The alphabet, for instance, may be recited with various types
of foot-division, e.g.
(i) || A | B | C | D | £ ||
(ii) || A B | C D | E F 1 G ||
(iii) || A B C | D | E F | G || etc.*
1.715 The elements of foot-structure are I (initial, or ictus ) and
R (reduced, or remiss)*. The exponent of I is always a single
syllable. The exponent of R, if present, is one, or more than one,
5 The feet and foot-divisions will be most apparent if the reader ‘beats time’
while reading these aloud, letting the down-beat coincide with the start of
each foot.
• The terms ictus and remiss have recently been revived by D. Abercrombie —
the first being a traditional term, the second used by Joshua Steele in Prosodia
Rationales (1779). They are used by M. A. K. Halliday in his ‘The Tones of
English’, Archivum Linguisticum, Vol. XV, Fasc. 1, pp. 1-28, 1964.
15
A LINGUISTIC THEORY OF TRANSLATION
syllable. Thus, in the examples above, the feet represented in
|| A | B | G | . . . etc. each have the structure I. Those repre-
sented by || A B | C D | . . . etc. have the structure IR with a
single syllable as exponent of R, while that represented by
|| A B C | has the same structure IR, but here R is expounded
by two syllables.
In some cases, the exponent of either I or R is a ‘zero syllable’
—that is, a momentary silence, or rest, represented by a caret ( A ) .
The time taken up by the rest is usually about that which is
needed to make up the duration of a full foot. When an utterance
begins with an unstressed syllable, we take this to be the exponent
of R in an initial foot, the exponent of I in this case being rest.
This appears to be justified by the fact that when such ‘incom-
plete’ feet occur immediately after a preceding utterance by the
same speaker there is commonly a momentary silence, which
makes up the time-lapse appropriate to a foot. Thus
(L I I didn’t | go there ||
1.716 Differences of foot-division are meaningful, being often
the exponents of differences in the structure of grammatical
units: e.g.
1. || John was a | light house | keeper ||
2. || John was a light | house keeper ||
Here the foot-division before ‘light’ in 1. marks light house as a
compound noun operating as exponent of H in the Ngp. The
foot-division between light and house in 2. is exponent of a
grammatical division, marking light as M in the Ngp, where
house is H.
1.717 The syllable. The syllable is the unit of syllabicity. Syl-
lables sometimes coincide with feet. When syllable-divisions occur
within a foot their phonic exponent is a momentary retardation
of the major chest-pulse movement.
The elements of syllable-structure are N (nucleus) and K
(consonantal, or marginal element) : the latter may be subdivided
as K r (consonantal syllable-releasing element), K a (consonantal
16
GENERAL LINGUISTIC THEORY
syllable-arresting element) and K' (consonantal inter-nuclear
unit 7 ). The unit K 1 occurs only between two Ns, and cannot be
assigned to either of them.
Syllable structures are thus: N, K r N, NK a , K r NK a , NK'(N),
(N)K*N. Examples: N oh! K r N tea, spar, straw, NK a at, and,
asks. K r NK a top, stop, stops, etc.
The exponents of N are V (simple vowel) or Vv (complex
vowel), the exponents of K r are C (one consonant) or CC or
CCC ; the exponents of K a are C, CC, CCC, CCCC.
1.718 The phoneme. Phonemes are the units of articulation
which operate as exponents of elements of syllable structure. The
primary classes are :
V, vowels — operating as exponent of N in syllable structure :
i e a o u a
v , glides — operating alone, or in complex vowels (V v ), as exponent
of N: iau
C, consonants: pbtdkgfv0Sszj3hmnglrwy
1.72 In English grammar we recognize a hierarchy of five units:
1. Sentence
2. Clause
3. Group
4. Word
5. Morpheme
1.721 Sentence : The primary elements of sentence-structure arc
a and (3. Sentence-structures which occur are a, [3, a [3, (3oc . . .
etc.
Examples: a. John arrived yesterday.
(3 When John arrived !
a(3 John arrived after we had left.
(3a After we had left, John arrived, etc.
The exponents of elements of sentence-structure are clauses.
1.722 Clause. The primary classes of clause are free (operating
as exponent of a in sentence-structure) and bound (operating as
exponent of (3 in sentence-structure).
* The interlude of C. F. Hockett Marmot of Phonology, p. 52.
17
A LINGUISTIC THEORY OF TRANSLATION
The primary elements of clause structure: S, P, C, A, have
been given above (1.33). Primary clause structures include:
SP e.g. hejcame they j had arrived etc.
SP (S inserted in P) e.g. did he come? had they arrived?
P (A etc.) e.g. Come! Come here.
The exponents of these elements are: P— one, or more than
one, Verbal Group (if more than one, the first is finite or non-
finite, the other (s) are non-finite), S and C — one, or more than
one, Nominal Group, A — one or more than one Adverbial Group.
1.723 Group. The primary group classes are Verbal , operating
at P in clause structure, Nominal, operating at S or C in clause
structure, and Adverbial , operating at A in clause structure.
Since practically no reference is made in the rest of this book
to the structure of groups other than Nominal, we confine our-
selves here to Nominal Groups.
We have already given the primary elements of Nominal Group
structure in 1.43 above: M, H and Q. The structures which
actually occur are:
H e.g. John, he, wine, etc.
M . . . H e.g. Old John, red wine, these three old books, etc.
HQ, e.g. John the Baptist, people who live in glass
houses, etc.
M . . . HQ e.g. the man in the moon, the old man who lives
next door, etc.
Secondary elements of Ngp structure, at M are d, o, and c
(already exemplified in 1 .43 above) .
The normal exponents of elements of group structure arc
words. In Ngps, however, we may have rank-shifted clauses and
rank-shifted groups as exponents, e.g. In What you say is wrong,
what you say is a rank-shifted clause (of structure CSP) operating
as exponent of H in the Ngp. What you say. In the Ngp the man
who came to dinner . . ., which has the structure MHQ, the
exponent of Q is the rank-shifted clause who came to dinner. In the
Ngp the man in the moon, which has the structure MHQ, the
exponent of Q is the rank-shifted Adverbial group in the moon.
1.724 Words. These fall into a large number of classes in terms
18
GENERAL LINGUISTIC THEORY
of their operation in the structure of groups. The primary ele-
ments of word-structure are B (base) and A (affix) . The exponents
of these are morphemes.
1.725 Morphemes. These are the smallest meaningful units of
grammar. They fall into two primary classes in terms of their
operation in the structure of words — base morphemes, and affix
morphemes. Since morphemes are at the ‘bottom’ of the rank scale
they themselves have no structure. In phonological and grapho-
logical exponence affix morphemes may be expounded linearly
(e.g. the Nominal plural morpheme expounded, most frequently,
by a suffixed graphological -s, or phonological / -s, -z, -*z /), or
exponentially fused with base morphemes (e.g. saw = fused
exponence of base morpheme SEE + affix morpheme ‘preterite’).
1.8 To conclude this introductory chapter we summarize the
field of linguistics and the linguistic sciences.
General Linguistics is the general theory of how language works.
It provides categories which are applicable in all branches of
linguistic science.
General Phonetics is the theory of phonic substance: it provides
categories which can be used in the description of the distinctive
phonic features of the phonological units of particular languages.
Descriptive Linguistics is the application and extension of general
linguistic categories in the description of particular languages.
Comparative Linguistics is an extension of descriptive linguistics
which establishes relations between two or more languages. When
the languages are separated in space, but not time, it is Syn-
chronic Comparative Linguistics, When they are separated in
time, it is Diachronic Comparative Linguistics.
Other parts of the general field of linguistics include Institu-
tional Linguistics and the theory of Language Varieties (dealt with
in Chapter 13).
Applied Linguistics is a term used to cover all those applications of
the theory and categories of general linguistics which go beyond
(i) the elucidation of how languages work and (ii) the description
of a particular language or languages*#* Us/jheir own sake. The
theory of translation is essentially^ the<^y/of>.pplied linguistics.
19
2
Translation: Definition and
General Types
2.0 The theory of translation is concerned with a certain type
of relation between languages and is consequently a branch of
Comparative Linguistics. From the point of view of translation
theory the distinction between synchronic and diachronic com-
parison is irrelevant. Translation equivalences may be set up,
and translations performed, between any pair of languages or
dialects — ‘related’ or ‘unrelated’ and with any kind of spatial,
temporal, social or other relationship between them.
Relations between languages can generally be regarded as two-
directional, though not always symmetrical. Translation, as a
process, is always uni-directional: it is always performed in a
given direction, ‘from’ a Source Language ‘into’ a Target Language.
Throughout this paper we make use of the abbreviations : SL =
Source Language, TL = Target Language.
2.1 Translation may be defined as follows:
the replacement of textual material in one language (SL) by equivalent
textual material in another language (TL).
This definition is intentionally wide — not vague, though it may
appear so at first sight. Two lexical items in it call for comment.
These are ‘textual material’ (where ‘text’ might have been
expected) and ‘equivalent’.
The use of the term ‘textual material’ underlines the fact that
in normal conditions it is not the entirety of a SL text which is
translated, that is, replaced by TL equivalents. At one or more
levels of language there may be simple replacement, by non-
equivalent TL material : for example, if we translate the English
text What time is it? into French as Quelle heure est-il? there is
replacement of SL (English) grammar and lexis by equivalent TL
(French) grammar and lexis. There is also replacement of SL
graphology by TL graphology — but the TL graphological form is
by no means a translation equivalent of the SL graphological form.
20
TRANSLATION : DEFINITION AND GENERAL TYPES
Moreover, at one or more levels there may be no replacement
at all, but simple transference of SL material into the TL text.
On this, see Chapter 6 below.
The term ‘equivalent’ is clearly a key term, and as such is
discussed at length below. The central problem of translation-
practice is that of finding TL translation equivalents. A central
task of translation theory is that of defining the nature and
conditions of translation equivalence.
Before going on to discuss the nature of translation equivalence
it will be useful to define some broad types or categories of
translation in terms of the extent (2.2), levels (2.3), and ranks (2.4)
of translation.
2.2 Full vs. Partial translation. This distinction relates to the
extent (in a syntagmatic sense) of SL text which is submitted to the
translation process. By text we mean any stretch of language,
spoken or written, which is under discussion. According to cir-
cumstances a text may thus be a whole library of books, a single
volume, a chapter, a paragraph, a sentence, a clause . . . etc.
It may also be a fragment not co-extensive with any formal
literary or linguistic unit.
2.21 In a full translation the entire text is submitted to the
translation process : that is, every part of the SL text is replaced
by TL text material.
2.22 In a partial translation, some part or parts of the SL text
are left untranslated: they are simply transferred to and incor-
porated in the TL text. In literary translation it is not uncommon
for some SL lexical items to be treated in this way, either because
they are regarded as ‘untranslatable’ or for the deliberate purpose
of introducing ‘local colour’ into the TL text. This process of
transferring SL lexical items into a TL text is more complex than
appears at first sight, and it is only approximately true to say
that they remain ‘untranslated’: on this, see 6.31.
2.23 The distinction between full and partial translation is
hardly a (linguistically) technical one. It is dealt with here,
however, since it is important to use the distinct term partial in
this semi-technical, syntagmatic, sense, reserving the term
restricted for use in the linguistically technical sense given in
2.3.
21
A LINGUISTIC THEORY OF TRANSLATION
2.3 Total vs. Restricted translation. This distinction relates to
the levels of language involved in translation.
2.31 By total translation we mean what is most usually meant
by ‘translation’ ; that is, translation in which all levels of the SL
text are replaced by TL material. Strictly speaking, ‘total’ trans-
lation is a misleading term, since, though total replacement is involved
it is not replacement by equivalents at all levels (cf. 2.1 above).
In ‘total’ translation SL grammar and lexis are replaced by
equivalent TL grammar and lexis. This replacement entails the
replacement of SL phonology/graphology by TL phonology/
graphology, but this is not normally replacement by TL equi-
valents , hence there is no translation, in our sense, at that level 1 . For
use as a technical term, Total Translation may best be defined as:
replacement of SL grammar and lexis by equivalent TL grammar and
lexis with consequential replacement of SL phonology I graphology by
( non-equivalent ) TL phonology / graphology .
2.32 By restricted translation we mean:
replacement of SL textual material by equivalent TL textual material, at
only one level,
that is translation performed only at the phonological or at the
graphological level, or at only one of the two levels of grammar
and lexis.
It should be noted that, though phonological or graphological
translation is possible, there can be no analogous ‘contextual
translation’ — that is translation restricted to the inter-level of
context but not entailing translation at the grammatical or
lexical levels. In other words there is no way in which we can
replace SL ‘contextual units’ by equivalent TL ‘contextual units’
without simultaneously replacing SL grammatical/lexical units by
equivalent TL grammatical/lexical units, since it is only by virtue
1 Occasionally there is concomitant replacement by a TL form which is
phonologically equivalent, or nearly equivalent, to the SL form at the phono-
logical level, as when Jap. He is translated by (Amer.) Eng. yeah, as it may be
in certain cases (see 5.6). When this happens in total translation it is normally
purely accidental. Rare cases of deliberate attempts at partial replacement by
equivalent TL phonology, in total translation, do occur: e.g. in film ‘dubbing’
and translation of poetry.
22
TRANSLATION : DEFINITION AND GENERAL TYPES
of their encapsulation, so to say, in formal linguistic units that
‘contextual units’ exist. Context is, in fact, the organization of
situation-substance into units which are co-extensive with and
operationally inseparable from the formal units of grammar and
lexis. With the medium levels the situation is different. Phono-
l°gy, for instance, is the organization of phonic substance into
units which, in combination, function as exponents of the units
of grammar and lexis ; phonological units, as such, are not bound
to grammatical or lexical units in the way in which contextual
units are bound to such units. Hence the separability of phono-
logy/graphology for translation purposes; and, on the other hand,
the non-separability of context.
2.321 In phonological translation SL phonology is replaced by
equivalent TL phonology, but there are no other replacements
except such grammatical or lexical changes as may result
accidentally from phonological translation : e.g. an English plural,
such as cats, may come out as apparently a singular cat in phono-
logical translation into a language which has no final consonant
clusters.
2.322 In graphological translation SL graphology is replaced by
equivalent TL graphology, with no other replacements, except,
again, accidental changes.
2.323 Phonological translation is practised deliberately by
actors and mimics who assume foreign or regional ‘accents’ —
though seldom in a self-conscious or fully consistent way (i.e.
except in the case of particularly good mimics, the phonological
translation is usually only partial). The phonetic/phonological
performance of foreign-language learners is another example of
(involuntary and often partial) phonological translation. Grapho-
logical translation is sometimes practised deliberately, for special
typographic effects, and also occurs involuntarily in the per-
formance of persons writing a foreign language.
Both phonological and graphological translation must be in-
cluded in a general theory of translation because they help to
throw light on the conditions of translation equivalence, and
hence on the more complex process of total translation.
2.324 Graphological translation must not be confused with
transliteration. The latter is a complex process involving phono-
23
A LINGUISTIC THEORY OF TRANSLATION
logical translation with the addition of phonology-graphology
correlation at both ends of the process, i.e. in SL and TL. In
transliteration, SL graphological units are first replaced by
corresponding SL phonological units; these SL phonological
units are translated into equivalent TL phonological units;
finally the TL phonological units are replaced by corresponding
TL graphological units. But the process is further complicated in
ways discussed in Chapter 10 below.
2.325 Restricted translation at the grammatical and lexical levels
means, respectively, replacement of SL grammar by equivalent
TL grammar, but with no replacement of lexis, and replacement
of SL lexis by equivalent TL lexis but with no replacement of
grammar. ‘Pure’ translation restricted to either of these levels is
difficult if not impossible owing to the close interrelations between
grammar and lexis and the tendency for exponents of grammatical
categories to be ‘fused’ with exponents of lexical items. Since the
grammatical categories of a language are relatively high-level
abstractions, ‘pure’ statements of grammatical equivalences can
best be presented as formulaic equations: but this is not trans-
lation, which is an operation performed on a specific SL text.
Grammatical translation requires that the SL text be replaced by
a text which is purely TL in its grammar, but still retains all the
SL lexical items. On this, see below.
2.4 Rank of Translation. A third type of differentiation in trans-
lation relates to the rank in a grammatical (or phonological)
hierarchy at which translation equivalence is established.
In normal total translation the grammatical units between
which translation equivalences are set up may be at any rank,
and in a long text the ranks at which translation equivalence
occur are constantly changing: at one point, the equivalence is
sentence-to-sentence, at another, group- to-group, at another
word-to-word, etc., not to mention formally ‘shifted’ or ‘skewed’
equivalences (see Chapter 12).
It is possible, however, to make a translation which is total in
the sense given in 2.31 above, but in which the selection of TL
equivalents is deliberately confined to one rank (or a few ranks,
low in the rank scale) in the hierarchy of grammatical units. We
may call this rank-bound translation . The cruder attempts at Machine
24
TRANSLATION : DEFINITION AND GENERAL TYPES
Translation are rank-bound in this sense, usually at word or
morpheme rank ; that is, they set up word-to-word or morpheme-
to-morpheme equivalences, but not equivalences between high-
rank units such as the group, clause or sentence. In contrast with
this, normal total translation in which equivalences shift freely up
and down the rank scale may be termed unbounded translation.
2.41 In rank-bound translation, as we have said, an attempt
is made always to select TL equivalents at the same rank, e.g.
word. A word-rank-bound translation is useful for certain pur-
poses, for instance, for illustrating in a crude way differences
between the SL and the TL in the structure of higher-rank units
— as in some kinds of interlinear translation of texts in ‘exotic’
languages. Often, however, rank-bound translation is ‘bad’ trans-
lation, in that it involves using TL equivalents which are not
appropriate to their location in the TL text, and which are not
justified by the interchangeability of SL and TL texts in one and
the same situation (see Chapter 7).
2.42 The popular terms free, literal, and word-for-word translation,
though loosely used, partly correlate with the distinctions dealt
with here. A free translation is always unbounded — equivalences
shunt up and down the rank scale, but tend to be at the higher
ranks — sometimes between larger units than the sentence. Word-
for-word translation generally means what its says : i.e. is essentially
rank-bound at word-rank (but may include some morpheme-
morpheme equivalences). Literal translation lies between these
extremes; it may start, as it were, from a word-for-word trans-
lation, but make changes in conformity with TL grammar (e.g.
inserting additional words, changing structures at any rank, etc.) ;
this may make it a group-group or clause-clause translation. One
notable point, however, is that literal translation, like word-for-
word, tends to remain lexically word-for-word, i.e. to use the
highest (unconditioned) probability lexical equivalent for each
lexical item. 2 Lexical adaptation to TL collocational or ‘idiomatic’
requirements seems to be characteristic oi free translation, as in
this example:
SL text It’s raining cats and dogs.
* On equivalance-probabilities, see 3.3 below.
25
A LINGUISTIC THEORY OF TRANSLATION
TL text 1 II est pleuvant chats et chiens. (Word-for-word)
2 II pleut des chats et des chiens. (Literal)
3 II pleut a verse. (Free)
Here 1 is word-word, 2 is group-group (with TL structural
‘normalizations’ within two of the groups). 3, since it changes the
clause-structure from SPC to SPA, must, perhaps, be regarded as
clause-clause: it also introduces a TL lexical normalization.
Only 3, the free translation, is interchangeable with the SL text
in situations.
Another example of free translation (switching to full sentence-
equivalence) would be this Russian-English one:
SL text Bogsn'im'i!
TL text 1 God with them ! (Word-for-word)
2 God is with them! (Literal)
3 Never mind about them! (Free)
Once again, only 3, the free translation, is interchangeable with
the SL text in a situation where the addressee is being advised to
dismiss or disregard a triviality.
26
3
Translation Equivalence
We have to distinguish between, on the one hand, translation
equivalence as an empirical phenomenon, discovered by com-
paring SL and TL texts ; and, on the other hand, the underlying
conditions, or justification, of translation equivalence. The con-
ditions of translation equivalence are discussed in Chapter 7.
Here we are concerned only with translation equivalence as an
empirical phenomenon.
3.1 A further distinction must be made between textual equi-
valence and formal correspondence. A textual equivalent is any TL
text or portion of text which is observed on a particular occasion, by
methods described below, to be the equivalent of a given SL text
or portion of text. A formal correspondent, on the other hand, is any
TL category (unit, class, structure, element of structure, etc.) which
can be said to occupy, as nearly as possible, the ‘same’ place in the
‘economy’ of the TL as the given SL category occupies in the SL.
Since every language is ultimately sui generis — its categories being
defined in terms of relations holding within the language itself— it
is clear that formal correspondence is nearly always approxi-
mate.
3.2 A textual translation equivalent, then, is any TL form (text or
portion of text) which is observed to be the equivalent of a given
SL form (text or portion of text).
3.21 The discovery of textual equivalents is based on the
authority of a competent bilingual informant or translator. Thus,
to find the French textual equivalent of the English text My son
is six, we ask a competent translator to put this into the TL,
French. He supplies Mon fils a six arts 1 . This, then, is the textual
equivalent of My son is six. We may repeat this process for any portion
of the full text — asking, for instance, for the French equivalent of
1 It should be noted that this, and almost all other examples in this paper,
are decontextualized texts: consequently the equivalents given are merely
probable (in this case highly probable). Some of them might be different in
special contexts.
27
A LINGUISTIC THEORY OF TRANSLATION
My son in this text. The translator supplies the equivalent
Mon JUs.
3.22 In place of asking for equivalents we may adopt a more
formal procedure, namely, commutation and observation of con-
comitant variation. In other words we may systematically intro-
duce changes into the SL text and observe what changes if any
occur in the TL text as a consequence. A textual translation
equivalent is thus : that portion of a TL text which is changed when and
only when a given portion of the SL text is changed. In our present
example, having had My son is six translated into French we
might ask for the translation of Tour daughter is six. The TL text
this time is Votre file a six ans. The changed portion of the TL
text (Mon fils/Votre fille) is then taken to be the equivalent of the
changed portion of the SL text (My son/Your daughter).
3.221 In simple cases like the above, one generally relies on
one’s own knowledge of the languages involved. This is the only
thing one can do with a recorded (spoken or written) text when
the original translator is not present. In such a case, the investiga-
tor acts as his own informant and discovers textual equivalents
‘intuitively’ — i.e. by drawing on his own experience, without
necessarily going through an overt procedure of commutation.
Nevertheless, commutation is the ultimate test for textual
equivalence, and it is useful in cases where equivalence is not of
the simple equal-rank and unit-to-unit type illustrated above.
3.222 For example, given the English SL text The woman
came out of the house, and its Russian TL equivalent Zenscina
vysla iz domu, we might wish to discover the Russian equivalent
of the English definite article in the group The woman in this text.
Commutation might give the following result:
SL text 1 The woman came out of the house.
TL text 1 Zenscina vy§la iz domu.
SL text 2 A woman came out of the house.
TL text 2 Iz domu vySla ienSiina.
We would thus establish that, in this particular position in this
particular text, the change of English the to a is correlated with a
change in the sequence of elements in the structure of the Russian
clause. We can state this textual equivalence as:
28
TRANSLATION EQUIVALENCE
Eng. the in (N) at /S/ = Rus. /SPA/
Eng. a in (N) at /S/ = Rus. /SPA/
This may be read as ‘English the, a term in a system operating in
a Nominal Group, at the place in Clause-structure, Subject, has
as its Russian translation equivalent the indicated sequence
(Subject, Predicator, Adjunct) of elements in the Russian clause
structure’, and, further, ‘English a, a term . . , etc., has its
translation equivalent, the inverse sequence of elements in the
Russian clause’.
3.223 In some cases there is no TL equivalent of a given SL
item, and commutation may again be used to demonstrate this.
It is useful to say in such cases that the TL equivalent is nil,
reserving the term zero for use, if at all, when zero is a term
operating in a TL system. Thus, to take another example, com-
paring the following English SL text and TL texts in French and
Russian, we see a possible use for the distinction between zero
and nil.
SL Eng. My father was a doctor.
TL Fr. Mon pere dtait docteur.
TL Rus. Otets u men 'a byl doktor.
One might describe the system of articles in both French and
English as containing a term zero. In the present example, then,
we could say that the translation equivalent of the English
indefinite article, a, is the French article zero. Russian, however,
has no system of articles. In the Russian text, therefore, there is
no translation equivalent of the English indefinite article. We
say, then, that the Russian equivalent of a in this text is nil.
Equivalence, in this example, can be established only at a higher
rank, namely the group. The English nominal group a doctor has
as its equivalent the Russian nominal group doktor. In general,
nil equivalence at one rank implies that equivalence can only be
established at a higher rank.
3.3 In a text of any length, some specific SL items are almost
certain to occur several times. At each occurrence there will be
29
A LINGUISTIC THEORY OF TRANSLATION
a specific TL textual equivalent. Having observed each particular
textual equivalent, we can then make a general statement of
textual equivalences for each SL item, covering all its occurrences
in the text as a whole. At each occurrence, the particular SL
item may always have the same TL equivalent. The statement
of general textual equivalence in this case is qualitatively the
same as that of particular textual equivalence; but there is a
difference, namely that it can be quantified. We may express it in
the actual figures, e.g. ‘SL item X occurs 79 times in this text,
and its TL equivalent is x in every case’; or as a percentage,
‘SL X = TL x, 100%’ ; or, finally, as a probability, in terms of the
probability scale in which 1 means ‘absolute certainty’ and 0
means ‘absolute impossibility’, ‘SL X = TL x, T, i.e. ‘SL X
has, as its textual equivalent, TL x, with the probability 1’. This
means that if you choose any occurrence of X in the SL text at
random, it is certain that its TL equivalent will be x.
3.31 Frequently occurring SL items commonly have more than
one TL equivalent in the course of a long text. Each particular
equivalent occurs a specific number of times: by dividing the
number of occurrences of each particular equivalent by the total
number of occurrences of the SL item we obtain the equivalence-
probability of each particular equivalence. For example, in a
French short story of about 12,000 words the preposition dans
occurs 134 times. The textual equivalent of this in an English
translation is in in 98 occurrences, into in 26, from in 2, and about
and inside in one occurrence each; there are six occurrences of
dans where the equivalent is either nil, or not an English pre-
position. (The short study from which these figures are taken did
not further differentiate these six cases.) In terms of probabilities
we can state the translation equivalences as follows: dans = in
■13, dans = into -19, dans = from '015, dans = aboutjinside -0075.
This means that if you select any occurrence of dans at random
in this text, the probability that its translation equivalent on that
occasion is in is -73, the probability that it is into is -19, etc.
3.32 The probability values given so far are based on the
assumption that, at each occurrence, the probability of a par-
ticular equivalence is the same as it is at every other occurrence ;
that is to say, they are unconditioned probabilities. But the equi-
30
TRANSLATION EQUIVALENCE
valence-probabilities are, in fact, constantly affected by con-
textual and co-textual factors 2 . We must, then, take these factors
into account, and consider not merely the unconditioned
probabilities, but also the conditioned probabilities of the various
equivalences. Thus, though the unconditioned probability of the
equivalence dans — into is only -19, the conditioned probability
of this equivalence is very much higher when dans is preceded by
certain verbs, e.g. aller, and must be 1 (certainty), or very nearly
so, when such a ‘verb of motion’ precedes, and a ‘noun referring
to a place’ follows.
Provided the sample is big enough, translation-equivalence-
probabilities may be generalized to form ‘translation rules’
applicable to other texts, and perhaps to the ‘language as a whole’
— or, more strictly, to all texts within the same variety of the
language (the same dialect, register, etc. — see Chapter 13).
3.4 A translation rule is thus an extrapolation of the probability
values of textual translation equivalences. Such a rule is a state-
ment of highest unconditioned probability equivalence, supple-
mented by highest conditioned probability equivalences, with an
indication of the conditioning factors. For human translators the
rules can make appeal to contextual meaning (e.g. ‘dans — translate
as in unless a verb of motion precedes and a place-noun follows’
or the like). For the purpose of Machine Translation, translation
rules may be operational instructions for co-textual search for items
marked in the machine glossary by particular diacritics, with
instructions to print out the particular conditioned equivalent in
each case. Such operational instructions, which if followed, can
be guaranteed with a high degree of probability to produce a
‘correct’ result, are known as algorithms. The looser, more con-
textually based, instructions for human translators are ‘trans-
lation rules’; the more rigid, co-textually based, instructions for
MT are, strictly speaking, ‘translation-algorithms’. In general, to
be effective, translation algorithms must be based on equivalences
with probabilities approaching 1 .
a By context we mean ‘context of situation’, i.e. those elements of the extra-
textual situation which are related to the text as being linguistically relevant:
hence contextual. By co-text we mean items in the text which accompany the item
under discussion : hence co-textual.
31
4
Formal Correspondence
In 3.1 above we alluded to the distinction between textual
equivalence and formal correspondence. A formal correspondent
is any TL category which may be said to occupy, as nearly as
possible, the ‘same’ place in the economy of the TL as the given
SL category occupies in the SL.
4.1 It is clear that formal correspondence can be only approx-
mate, and that it can be most easily established at relatively high
levels of abstraction. Thus, if we find that two languages operate
each with grammatical units at five ranks (an example might be
English and French, both of which appear to have five ranks:
sentence, clause, group, word, morpheme) we can reasonably say
that there is formal correspondence between the two hierarchies
of units; each has the same number of ranks, and as (taxonomic)
hierarchies each has the same kind of relationship between units
of the different ranks. Having established such a highly abstract
correspondence, we may use this as a frame of reference for
stating approximate correspondence at lower abstractional levels;
e.g. we may talk of formal correspondence between SL and TL
elements of structure operating at ‘corresponding’ ranks.
4.2 It may be, however, that formal correspondence can only
be established ultimately on the basis of textual equivalence at
some point. Thus we may state that an item or class of one
language is the formal equivalent of an item or class in another,
because the category in question operates in approximately the
same way in the structure of higher rank units in both languages;
but this in turn, implies that we have established a correspondence
between these higher rank units, and this may have to be done
on the basis of highest probability textual equivalence.
4.21 For example, we might say that there is formal corres-
pondence between the word-classes preposition in English and
French. This statement is based on the fact that in both languages
the word-class labelled ‘preposition’ functions along with nominal
groups in the structure of adverbial phrases, which in turn
32
FORMAL CORRESPONDENCE
function in both languages as (i) qualifiers in nominal group
structure (e.g. the door of the house — la porte de la maison) or
(ii) as adjuncts in clause structure. But this clearly pushes the
problem of justifying our statement of formal correspondence
further up the rank scale; we still have to justify the correspond-
ence of nominal groups, adjuncts, etc., and this might have to be
done on the basis of textual equivalence.
4.3 In spite of its approximate nature, and the theoretical
difficulty ofits justification, the concept of ‘formal correspondence’
is a useful one; indeed, it is an essential basis for the discussion of
problems which are important to translation theory and necessary
for its application (see Chapter 12).
4.31 Formal correspondence is of interest from another point
of view as well; namely that the degree of divergence between
textual equivalence and formal correspondence may perhaps be
used as a measure of typological difference between languages.
This can be exemplified by considering formal correspondence
and textual equivalence between English prepositions and - certain
formal classes in French and Kabardian (a N.W. Caucasian
language).
4.311 In the French text referred to above there are 1220
occurrences of prepositions. In the English TL text 910 of these
have a preposidon as textual translation equivalent: for this
text, the unconditioned equivalence-probability of the equivalence
Fr. preposition = Eng. preposition is -75. We are justified in saying
that for English and French prepositions there is a fairly high
degree of convergence between formal correspondence and
textual equivalence; and this may be taken as a symptom of
typological similarity.
4.312 The establishment of formal correspondences between
English and Kabardian is more difficult; for one thing, it is
probable that Kabardian has only four ranks of grammatical
units as compared with the five of English. We may, however,
roughly equate units of the lowest rank in both languages,
labelling both morphemes. In Kabardian there is a class of bound
morphemes which may be called ‘relational preverbs\ These are
prefixed to verbal morphemes, forming together with them (and
certain other morphemes) verbal units which funcdon as predi-
33
A LINGUISTIC THEORY OF TRANSLATION
cators in clause structure. Formally, it is reasonable to say that
these relational preverbs correspond most closely to English
bound morphemes such as in- ex- etc., which occur prefixed to
verbs; in other words, Kabardian relational preverbs are formal
correspondents of English verbal prefixes. No actual figures for
textual equivalence are available, but it is almost certain that
the highest-probability English textual equivalents of Kabardian
relational preverbs are prepositions. There is thus considerable
divergence between formal correspondence and textual equi-
valence as between English prepositions and Kabardian relational
preverbs. This is what one might expect in the case of a pair of
languages which are both typologically and genetically very
different; more precisely, the divergence shown here may be
taken to be a symptom of typological difference, which parallels
genetic unrelatedness.
34
5
Meaning and Total Translation
It is generally agreed that meaning is important in translation —
particularly in total translation. Indeed, translation has often
been defined with reference to meaning; a translation is said to
‘have the same meaning’ as the original. Dostert defines trans-
lation as ‘that branch of the applied science of language which is
specifically concerned with the problem — or the fact — of the
transference of meaning from one set of patterned symbols . . .
into another set of patterned symbols . . .’ l
It is clearly necessary for translation-theory to draw upon a
theory of meaning; without such a theory certain important
aspects of the translation process cannot be discussed, nor can
statements like that of Dostert be evaluated. In terms of the
theory of meaning which we make use of here — a theory deriving
largely from the views of J. R. Eirth— the view that SL and TL
texts ‘have the same meaning’ or that ‘transference of meaning’
occurs in translation is untenable.
5.1 Meaning, in our view, is a property of a language. An SL
text has an SL meaning, and a TL text has a TL meaning — a
Russian text, for instance, has a Russian meaning (as well as
Russian phonology/graphology, grammar and lexis), and an
equivalent English text has an English meaning. This is neces-
sarily the case, since, following Firth, we define meaning as the
total network of relations entered into by any linguistic form —
text, item-in-text, structure, element of structure, class, term in
system— or whatever it may be.
The relations entered into by the formal linguistic units of
grammar and lexis are of two kinds (i) formal relations, (ii)
contextual relations.
5.11 By formal relations we mean relations between one formal
item and others in the same language. In grammar this may be the
relation between units of different rank in the grammatical hier-
1 Locke and Booth Machine Translation of Languages (New York, London
1955), p. 124.
35
A LINGUISTIC THEORY OF TRANSLATION
archy, the relation between terms in a system, the relation
between a class and an element of structure at a higher rank,
co-textual relations between grammatical classes or items in a
text, and so on. In lexis there are formal relations between one
lexical item and others in the same lexical set, and formal
co-textual (collocational) relations between lexical items in texts.
The various formal relations into which a form enters constitute
its formal meaning.
5.12 By contextual relations we mean the relationship of gram-
matical or lexical items to linguistically relevant elements in the
situations in which the items operate as, or in, texts. Those
situational elements which are contextually ‘relevant’ to a given
grammatical or lexical item are discovered, just as translation
equivalents are discovered, by commutation. Change an element
in the situation and observe what textual change occurs; change
an item in a text and observe what situational changes occur.
The range of situational elements which are thus found to be
relevant to a given linguistic form constitute the contextual
meaning of that form.
5.2 Now since every language is formally sui generis and formal
correspondence is, at best, a rough approximation, it is clear that
the formal meanings of SL items and TL items can rarely be the
same. A TL dual may on occasion be the translation equivalent
of an SL plural — for instance, Arabic kitaabeen as equivalent of
English books — but it cannot have the same formal meaning. One
is a term in a 2-term number-system, the other a term in a 3-term
system; each gets a ‘value’ 2 deriving from the co-existence of the
other term(s) in the system. We cannot, therefore, talk about
formal meaning being ‘transferred’ from SL to TL.
The same is true of contextual meanings. The contextual
meaning of an item is the groupment of relevant situational
features with which it is related. This groupment varies from one
language to another. It is rarely the same in any two languages,
and it is, moreover, related to formal meaning; thus, if we have
two systems containing different numbers of terms (and hence
differing as to the formal meanings of these terms) we will find that
2 In the Saussurean sense : this forma! ‘value’ of terms in a formal system is
also what is quantified as ‘information’ in information theory.
36
MEANING AND TOTAL TRANSLATION
at least some of the terms also have different contextual meanings.
5.3 Consider, for instance, two different systems of deictics or
demonstratives: one, a three-term system found in N.E. Scots
dialects (this-that-yon), the other, the four-term system of
Standard English (this-these-that-those) . If we assume that both
systems cover approximately the same total contextual field we
can represent the contextual meanings of the constituent terms
diagrammatically as follows :
S P
this
these
that
those
St. E.
this
that
yon
N.E. Sco.
1
2
3
The Standard English system is represented here as con-
textually 2-dimensional: it embodies two degrees of deixis (I, II)
and two numbers (Singular, Plural). The N.E. Scots system is
unidimensional, embodying only deixis — 3 degrees this time
(l, 2, 3). Numerosity is a contextually irrelevant feature of
situations for the N.E. Scots system.
It is clear that if we translate from Standard English to Scots we
cannot ‘transfer meaning’. There is no way in which, for example,
Scots that can be said to ‘mean the same’ as English that or this
or these or those. On a given occasion it may refer to, or be rela-
table to, the same feature of the situation as one of the English
deictics — but its formal and contextual meaning is clearly different.
5.4 A more extended example will make this point clearer.
Imagine a situation in which a girl walks in and says ‘I’ve
arrived’. The situation in which this text occurs is, like all
situations, indefinitely complex, in the sense that in an attempt
to describe it exhaustively one could go on sub-describing with
greater and greater delicacy to a degree ultimately limited only
by the refinement of our language of description. We might begin,
for example, by specifying the precise time, date and location, the
girl’s name, age, height, weight, colour of eyes and hair, her
37
A LINGUISTIC THEORY OF TRANSLATION
clothes, profession, religion, relationships to other people; the
number and nature of her audience . . . and so on.
5.41 However, only very few features of the situation are
linguistically relevant ; that is to say, are built into the contextual
meaning of the text and its parts. These include the following :
(i) one participant in the situation, identified only as the per-
former of the linguistic act (in this case, the speaker ) and correlated
with selection of the pronoun / as opposed to we, you, he, etc., or
a noun such as Mary.
(ii) an arrival — a complex event which need not be further
described here, correlated with selection of the lexical item
ARRIVE as opposed to, say, LEAVE or EAT, etc.
(iii) a prior event which is
(iv) linked to a current situation; these two together being cor-
related with selection of a perfect form ( have arrived ) as opposed to
a non-perfect ( arrive , arrived, etc.).
(v) the ‘current situation’ in this case is present, and this cor-
relates with selection of non-preterite ( have arrived) as opposed to
preterite ( had arrived) . . . etc. 3
5.42 Now suppose the text is translated into Russian as ja
prisla.
The relevant features of the situation now include:
(i) a speaker — selection ofja (as opposed to my, ty, etc.)
(ii) the speaker is female — selection of prisla (as opposed to prise!)
(iii) an arrival — selection of PRIITI (as opposed to VYITI, etc.)
(iv) on foot — selection of PRIITI (as opposed to PRIEXAT')
(v) a prior event — selection of past (as opposed to present, etc.)
(vi) completed on a specific occasion — selection of perfective (deter-
minate) verb-form (as opposed to imperfective) . . . etc.
5.43 Clearly, though the Russian text is a perfectly good
translation-equivalent of the English text, it does not ‘mean the
same’ — since it selects as linguistically (contextually) relevant a
different set of elements in the situation. We can tabulate the
difference thus:
» This list could be extended to cover the contextual meanings of the
Sentence, the Clause-Structure (SP), of Nominal Group, Verbal Group, etc., but the
list given here is sufficient for our purpose.
38
MEANING AND TOTAL TRANSLATION
I
have
arrived
speaker ■
female
arrival
on foot
prior event
linked to
present
completed
J a
priSla
Only the situational features italicized in the list are con-
textually relevant to both the SL and the TL text.
Such examples could be multiplied indefinitely. We will give
only two further illustrations at this point.
5.5 A Burushin 4 * is talking about his brother. In the course of
the conversation he frequently uses the item a-cho. The interpreter
translates this my brother. The Burushin is now replaced by his
sister. She, too, talks about the same person, who is, of course,
also her brother; she says a-yas & . The interpreter translates as
before : my brother.
5.51 Now, an Englishman might say ‘The interpreter trans-
lated a-cho by my brother. He is a good interpreter, so we may
assume that “my brother” means the same as “a-cho” ’. But, on
the second occasion, the interpreter translates a-yas by my brother.
So the Englishman — still trusting the interpreter — has to admit
that my brother ‘means the same as’ a-yas. Further pressed, he
asserts that my brother ‘means the same’ on both occasions. Now,
since the first ‘my brother’ means the same as the second ‘my
brother’ he must conclude that the two Burushaski items which
‘mean the same’ as ‘my brother’ also mean the same as each other.
At this point the Englishman, being an explorer or a travelling
salesman, undismayed by the unlikelihood of free variation,
refuses to discuss the matter further. The linguist, however,
cannot let the matter rest there. Unless a-cho and a-yas are free
variants, then they cannot ‘mean the same’ as each other. It is
4 That is, a speaker of Burushaski, the language of the Nagir and Hunza
States in the Gilgit Agency, in the extreme N.W. of Pakistan.
6 The a- in these examples is an (obligatory) pronominal prefix — the lexical
items are cho and y as.
39
A LINGUISTIC THEORY OF TRANSLATION
clear, then, that neither of them can mean the same as my brother;
for my brother, as the Englishman said, ‘means the same’ both
times. This is certainly true from an English point of view, and
this is the only linguistically valid point of view, since my brother,
being an English item, has an English meaning. It is equally true
that a-cho and a-yas, though they seem to ‘mean the same’ from
an English point of view, do not mean the same from a Buru-
shaski point of view; and this, again, is the only linguistically
valid point of view, since, being Burushaski items, they have
Burushaski meanings.
5.52 In fact, of course, brother and cho do not ‘mean the same’.
There is no ‘transference of meaning’ here; only replacement of
Burushaski items by an English translation equivalent. The situa-
tional elements which are, so to speak, encapsulated in the
contextual meaning of brother might be roughly characterized as
male and sibling: those which are encapsulated in the contextual
meaning of cho are sibling and of same sex as speaker. The relation-
ship of the English and Burushaski lexical items to elements in
the situation can be tabulated as follows : (in this tablfc + means
male, — means female )
English
Situation
Burushaski
Speaker
Sibling
brother
+
+
cho
—
+
yas
sister
+
—
cho
5.6 By a curious coincidence we can diagrammatize the con-
textual meaning of terms in the (grammatical) closed system of
‘acceptance-rejection’ items (yes-no) in exactly the same way.
In English, selection of^ics or no in response to a question (or
statement) depends on what we may call ‘the polarity of the
situation’: situation positive, answer ‘yes’; situation negative,
40
MEANING AND TOTAL TRANSLATION
answer ‘no’ (irrespective of the polarity of the preceding utter-
ance). In many other languages (e.g. at least some Slavonic
languages, Arabic, Japanese, at least some Bantu languages),
selection of the appropriate response depends on the polarity-
relationship between question (or statement) and situation:
same polarity — answer X; different polarity — answer T. Some lan-
guages (e.g. French, with oui, si, non, and Norwegian and Swedish,
with ja,jo, nej) have a three-term system here.
We illustrate with English, Japanese and French:
Question
Situation
Eng.
jap-
Fr.
. Did you ? I did.
+
+
yes
hai
oui
Didn’t you? I did.
—
+
iie
si
Did you ? I didn’t.
+
—
no
non
Didn’t you? I didn’t.
—
—
hai
We cannot say that English yes ‘means the same’ as Japanese
hai or French oui, though in certain situations hai and oui may be
the appropriate translation equivalents oiyes.
5.7 Another manifestation of the ‘same-meaning’ or ‘meaning-
transference’ fallacy is seen in the view that translation is a
‘transcoding’ process, a well-known example being Weaver’s
remark: ‘When I look at an article in Russian, I say: “This is
really written in English, but it has been coded in some strange
symbols. I will now proceed to decode”.’ 6
This implies either that there is a one-to-one relationship
between English and Russian grammatical/lexical items and their
contextual meanings, or that there is some pre-existent ‘message’
with an independent meaning of its own which can be presented
or expounded now in one ‘code’ (Russian) now in another ‘code’
(English). But this is to ignore the fact that each ‘code’ (i.e. each
language) carries with it its own particular meaning, since
meaning, as we have said in 5.1 above, is ‘a property of a
« In Locke and Booth Machine Translation of Languages (New York, London
1955), p. 18.
41
A LINGUISTIC THEORY OF TRANSLATION
language’. The one thing which does most nearly correspond to
transcoding is the universal literate practice of switching from
the spoken to the written medium and vice versa. In this process,
one and the same ‘message’ embodied in a particular selection of
grammatical and lexical forms may be presented, or expounded,
alternatively in two different ‘codes’. Thus the ‘message’ dis-
cussed in 5.4 above may be expounded in the written medium as
I've arrived and in the spoken medium in a phonological form
represented here in the transcription JJ 1. | ra'vd || . The
passage from one to the other of these two exponential forms of
the message might legitimately be called ‘transcoding’ — but this
is not translation.
It is possible that the ‘transcoding’ view is an operationally
useful one for machine translation, in its cruder forms at least,
where the problem is that of setting up algorithms which will
produce moderately intelligible translations with a high degree
of probability. But for tne deeper understanding of the translation
process the ‘transcoding’ view is not useful.
5.8 Our objection to ‘transcoding’ or ‘transference of meaning’
is not a mere terminological quibble. There are two main reasons
why translation theory cannot operate with the ‘transference of
meaning’ idea. In the first place, it is a misrepresentation of the
process, and consequently renders the discussion of the conditions
of translation equivalence difficult; in the second place, it con-
ceals the fact that a useful distinction can be made between
translation and another process which we call transference. In
transference, which we discuss in the next section, there is, indeed,
transference of meaning ; but this is not translation in the usual
sense.
42
6
Transference
In normal translation, as we have said above, the TL text has
a TL meaning. That is to say, the ‘values’ of TL items are entirely
those set up by formal and contextual relations in the TL itself.
There is no carry-over into the TL of values set up by formal or
contextual relations in the SL.
It is, however, possible to carry out an operation in which the
TL text, or, rather, parts of the TL text, do have values set up in
the SL\ in other words, have SL meanings. We call this process
transference.
6.1 A good example of transference of the formal and con-
textual meanings of lexical items is found in an article in Language
on colour terms in Navaho 1 . We will say more about Navaho
colour terms later (see 7.4 1 below) : meanwhile we consider only
two'f'ico and dootl’iz. These terms belong in a lexical set of only
three terms covering approximately the whole spectrum — -a situ-
ational range covered in English by six terms (red, orange, yel-
low, green, blue, purple). 2 Consequently, the formal meanings of
the Navaho terms must be different from those of the English
terms. The contextual meanings of the two Navaho terms given
here are also different from anything in English :4 ico means ap-
proximately the same zs, yellow + orange, and dootl'iz. as green +
blue + purple.
6.11 Now, for the purpose of ‘translation’, the authors of the
article coin two new ‘English’ colour terms: (it is) yoo (‘yellow-
or-orange’) as the ‘translation equivalent’ of4ico, and (it is)
bogop (‘blue-or-green-or-purple’) as the ‘translation equivalent’
1 Landar, Ervin and Horowitz, ‘Navaho Color Categories’, Language, 36.3(1)
(1960), p. 368. For typographic reasons some slight transcriptional alterations
have been made here.
2 Or seven if we use the time-honoured, but slightly technical, series
ROYGBIV, with indigo and violet. The whole treatment of colour and colour-
relations is somewhat over-simplified here and in 7.41 . This does not, however,
in any way affect the principle under discussion.
43
A LINGUISTIC THEORY OF TRANSLATION
of dootl’iz. We put the term ‘translation equivalent’ in quotes,
because here we have, not a case of translation, but of transference.
Among utterances recorded by Landar et al. we have:
(i) c'il bit’aa? nahalq dootl'iz: ‘it is bogop like a plant leaf*
(ii) k'os nahalo dootl’iz: ‘it is bogop like the sky’
(iii) c4dfd44h nahalo dootl'iz: ‘it is bogop like “purple-four-
o’clock” (a flower)’
It is clear (from co-textual evidence) that the appropriate
translation equivalents of dootl'iz would be (i) it is green, (ii) it is
blue, and (iii) it is purple. These are English lexical items — with
formal meanings derived from their membership of an English
lexical set, and contextual meanings likewise delimited by the
contextual meanings of the other members of the English set. In
place of these the form bogop is used: this is phonologically and
graphologically English — but insofar as it has any formal and
contextual meaning, this is derived from membership of a lexical
set in Navaho. Bogop, is, in fact, the Navaho item dootl’iz fitted
up with English phonological/graphological exponents (derived
acrophonically from the graphological exponents of the highest-
probability English translation equivalents).
6.2 Transference can also be carried out at the level of gram-
mar. In grammatical transference SL grammatical items are
represented in the TL text by quasi-TL grammatical items
deriving their formal and contextual meanings from the systems
and structures of the SL, not the TL.
6.21 As an example, we take Indonesian pronouns. Excluding
optional (or style-linked) plurals and certain nouns which may
function translationwise as equivalents of English pronouns (e.g.
the polite tuan and njonja, and the politically conscious saudara
‘brother’), Bahasa Indonesia has a nine-term pronoun-system,
as opposed to the English seven-term system. Evidently, then,
the terms in the two systems have different formal meanings;
they have, in addition, different contextual meanings. These
differences are approximately indicated in the system-diagrams
on the next page.
By superimposing these two system-diagrams we get an indica-
tion of translation equivalences: aku + saja = I, Kami + kita =
44
TRANSFERENCE
Indonesian English
S p S P
aku /
/ saja
(ex.)
kami
(in.)
kita
I
I
We
engkau /
/ kamu
you
ia z'
/ beliau
mereka
he
they
she
it
we, etc. The system-derived ‘values’, or meanings — formal and
contextual — of the items are different ; the Indonesian system, for
instance, contains two dimensions absent from the English system:
exclusive {inclusive (karni/kita) and familiar (non-familiar (aku/saja,
engkau/kamu, ia/beliau — though in this case the ‘non-familiar’
term, beliau, is distinctly honorific). The English system has a
gender dimension {hejshe(it), absent in Indonesian; and the ex-
tension of it and they ‘downwards’, beyond the formal-contextual
range of Indonesian pronouns should also be noted.
These diagrams give an approximate indication, as we have
said, of translation equivalence; at the same time they indicate
that not one English translation equivalent has ‘the same
meaning’ (formally or contextually) as any Indonesian pronoun.
6.22 We can, however, create a quasi-English set of transference-
equivalents : that is, of items with the formal and contextual
meanings of the terms in the Indonesian system. This can be done
in several ways: we might press into service the rare or archaic
English thou, as transference of engkau, we might use diacritic
letters or numbers, or we might modify the graphological form
of English items by adding mnemonically chosen letters: e.g.
45
A LINGUISTIC THEORY OF TRANSLATION
-h (non-familiar or (slightly) Aonorific), -e (exclusive), -i (inclu-
sive). We adopt the last procedure. We can now set up a system
of quasi-English pronouns which have (because we have so
decided) exactly the same meanings as the Indonesian pronouns 3 .
These items are in fact, formally and contextually, the Indo-
nesian pronouns; but they are fitted up with English grapho-
logical exponents, derived (with arbitrary modifications) from
the graphological exponents of the highest-probability English
translation equivalents. The transference-equivalents could be
embedded in English translations of Indonesian texts and might
conceivably be a useful device in teaching the Indonesian use of
Indonesian pronouns to English learners; for this purpose the
translation might be rank-bound at word rank, true translation
equivalents (with English meanings) being used for everything
except the pronouns.
6.3 In ‘real life’ transference is not very common. At first sight,
it seems as if the use by a translator of an SL lexical item em-
bedded in a TL text is pure transference. Yet reflection shows
(and, in theory, this could be experimentally verified) that an
SL lexical item in these circumstances does not fully retain its
SL meaning.
* Additional quasi-English transference-forms could be found for the bound
forms of such Ind. pronouns as have additional forms: e.g. ku = me, -nja - him,
possibly -mu = youhm, etc.
46
TRANSFERENCE
6.31 We might, for example, imagine the translator of a
Finnish novel rendering the sentence Menen saunaan into English
as I'm going to the sauna. Here, the lexical item sauna appears to
have been transferred bodily into the TL. Has it, however, taken
with it the meaning which it has in Finnish? Certainly not its
formal grammatical meaning: probably not its formal lexical
meaning nor all of its contextual meaning. For the translator
himself, knowing Finnish, it may be that it has its full Finnish
meaning; for an English reader it carries a contextual meaning
something like ‘foreign — specifically Finnish — cultural object or
institution somewhat comparable, as the contextual meanings of
the co-text show, with the Turkish Bath’ — and it is immediately
formalized as a (foreign) member of lexical sets containing items
like bath, steambath, Turkish Bath, Public Baths . . . etc.
6.32 An even better example is the lexical item sputnik, which
first occurred in English as a ‘transference’-item in October 1957.
In the co-texts in which it appeared it had the meaning of
‘(Russian) artificial satellite’ — no more. In Russian, of course, the
lexical item sputnik is a member of a number of lexical sets, and
would have an appropriate highest probability English trans-
lation equivalent in each: e.g.
fellow-traveller (traveller, wayfarer, companion . . . etc.)
companion (guide to, handbook, introduction)
satellite (planet, earth, moon . . . etc.)
( artificial ) satellite (space-ship, rocket . . . etc.)
In English, however, this item was introduced, and has re-
mained, within only the last lexical set, and with the appropriate
contextual meaning. Clearly, then, embedded in an English TL
text, or, now, simply in an English text, sputnik has an English
formal and contextual meaning. Since, however, this English
meaning of sputnik correlates with part of the total formal-
contextual meaning of Russian sputnik we may perhaps speak of
partial transference of meaning.
6.321 In fact, in the case of sputnik four different processes
are involved :
(i) lexical {partial ) transference.
(ii) grammatical translation: ‘sputnik’ as exponent of the Russian
47
A LINGUISTIC THEORY OF TRANSLATION
word-class ‘noun’ is replaced by ‘sputnik’ as exponent of the
equivalent, but not identical, English word-class ‘noun’.
(iii) phonological translation: the Russian phonological form
/ sputn'ik / is replaced by the equivalent English phonological
form / sputnik / or / spu u tnik / or, with a graphologically
derived adaptation, / spatnik /.
(iv) graphological transliteration: the Russian form cnymnuK is
replaced not by the English graphological translation equivalent
cnymhuk, but by the transliteration-equivalent sputnik 4 5 .
6.4 Pure meaning-transference may occur when a TL text
contains a TL word in its normal TL phonological/graphological
form, but with a contextual meaning taken over from the SL.
This may happen when one is speaking a foreign language. The
Russian who says My foot hurts, when he means ‘My leg hurts’ is
using purely English exponents, but transferring the contextual
meaning of Russian noga into an English text. Transferences of this
kind occur, though in what is not strictly a translation-situation,
in Indian English — for instance in novels about India written in
English by Indians. Examples® are government used, like Hindi
sarkar, to mean not only the institution of government but also
(sp. as a term of address) a person who represents government;
flower-bed, used by B. Bhattacharya, like Bengali phul-shajja, to
mean ‘nuptial bed’ ; brother, used like Hindi bhai both as a kinship
term and as a term of address and affection.
6.5 From examples like the foregoing it should be clear that a
restricted kind of ‘transference of meaning’ from one language to
another is possible ; but it is equally clear that this is not what is
normally meant by ‘translation’. In translation, there is substitution
of TL meanings for SL meanings : not transference of SL mean-
ings into the TL. In transference there is an implantation of SL
meanings into the TL text. These two processes must be clearly
differentiated in any theory of translation.
4 It sometimes seems a pity that we do not practise graphological translation.
Cnymhuk, presumably pronounced / nimhuk /, would be phonaesthetically
appropriate !
5 From B. B. Kachru An Analysis of Some Features of Indian English: a study
in linguistic method. (Ph.D. thesis, Edinburgh University, 1962.)
48
7
Conditions of Translation
Equivalence
7.1 We are now in a position to consider the necessary condi-
tions in which a given TL item can, or does, function as trans-
lation equivalent of a given SL item.
The SL and TL items rarely have ‘the same meaning’ in the
linguistic sense; but they can function in the same situation. In
total translation, SL and TL texts or items are translation
equivalents when they are interchangeable in a given situation. This
is why translation equivalence can nearly always be established
at sentence-rank — the sentence is the grammatical unit most
directly related to speech-function within a situation.
7.2 As our examples in Chapter 5 showed, in total translation
SL and TL items have overlapping meanings; their contextual
meanings include relationship to certain situational features in
common. In the case of Eng. 1 have arrived/ Russ, ja prisla we saw
that even for the rough characterization given in 5.4 we had to
specify 8 situational features: 5 for the English text, 6 for the
Russian. Only three of these (a speaker, an arrival and a prior event)
were common to both. The TL text must be relatable to at least
some of the situational features to which the SL text is relatable.
Presumably, the greater the number of situational features
common to the contextual meanings of both SL and TL text,
the ‘better’ the translation. The aim in total translation must
therefore be to select TL equivalents not with ‘the same meaning’
as the SL items, but with the greatest possible overlap of situa-
tional range. We will return later to the special problems which
arise when the situation contains elements relevant to the SL
’text, but absent from the cultural context of the TL.
7.3 In order to generalize our statement of the conditions of
translation equivalence so as to be applicable to restricted trans-
lation as well as total translation we must examine these ‘situa-
tional features’ or elements more closely.
49
A LINGUISTIC THEORY OF TRANSLATION
7.31 The bundles of situational features which are contextually
relevant to a text — that is, those which determine the selection of
this or that linguistic form as opposed to any other — are bundles
of distinctive features', and these are quite analogous to distinctive
features in phonology.
Thus, the situational features arrival, prior event, linked to,
present are situational distinctive features which distinguish the
contextual meaning of have arrived from that of have left, or arrive ,
or arrived, or had arrived, in much the same way as stop, labial,
voiceless, oral are distinctive features which distinguish the English
phonological unit / p / from / f /, / 1 /, / b /, / m /.
7.32 Now, the distinctive features of phonology are, in fact,
features of phonic substance, categorized in general phonetic terms;
general phonetics being the theory of phonic substance from which
we derive descriptive categories (‘labial’, ‘voiceless’, etc.) which
can be used for describing the distinctive features of phonological
units of particular languages. There is, as yet, no general theory
of situation-substance, no general semetics (or general pleretics) 1
from which to draw descriptive terms for the distinctive features
of contextual meanings of grammatical or lexical items in
particular languages.
We are therefore forced to operate with ad hoc terms in dis-
cussing contextual meaning and its relation to situation-substance.
But the parallelism holds good; the distinctive features of phono-
logy are phonetically categorized features of phonic substance,
the distinctive features of contextual meaning are (semetically
categorized) features of situation substance.
7.4 It is now possible to generalize the conditions for trans-
lation equivalence as follows:
translation equivalence occurs when an SL and a TL text or item are
relatable to (at least some of) the same features of substance . 2
1 The obvious choice of term is an -etic derivate either of sem- (from Gk. sema)
as in many commonly used terms, or a derivate of pier- (from Gk. pleres), as
used in Glossematics.
1 The type oi substance depends on the scope of the translation. For total
translation it is situation-substance: for phonological translation it is phonic-
substance: for graphological translation it is graphic-substance.
50
CONDITIONS OF TRANSLATION EQUIVALENCE
7.41 We can illustrate this from examples already given. The
Navaho colour terms referred to in 6.1 above were4ico and
dootl'iz. If we a dd4icu : we have, very roughly, complete cover-
age of the visible spectrum. Using the set of English colour terms,
Red, Orange, Yellow, Green, Blue, Purple, we can set up an
approximate translation-equivalence diagram as follows:
Form:
R
O
Y
G
B
p
Substance :
X
y
z
Form:
4icll ?
4ico
dootl’i¥
English
The spectrum
Navaho
The visible spectrum is a continuum of situation-substance.
This substantial continuum is dissected and organized into the
contextual meanings of English and Navaho linguistic forms
roughly as indicated — though, of course, with much less clear-
cut divisions than are shown here. As we showed in 6.1 the
Navaho terms4ico and dootl’iz (and no doubt also4 ici( J ) do not
have the same contextual meanings as any English terms. They
are, however, relatable to stretches of the same situation-sub-
stance. Let x, y and z represent actual colours present in particu-
lar situations, and relatable to the English terms red, green, blue
occurring in English texts in these situations. The Navaho terms
+ icn and dootl’iz function as translation-equivalents in these
situations because they are relatable to the same substantial fea-
tures x, y and z. It is solely this relationship to the same sub-
stantial features that justifies their use as translation-equivalents
since they clearly have quite different meanings from the English,
items.
7.42 In the same way at the level of grammar, English we
and Bahasa Indonesia kami are translation equivalents in a
situation where an English speaker excludes the addressee:
e.g. If you do this, WE will do that. Here the actual situational
features relatable to WE are the speaker (S) and at least one
other (O) ; the addressee (A) is excluded. Hence the translation
equivalence:
51
A LINGUISTIC THEORY OF TRANSLATION
where the outer line in each case indicates the total (potential)
system-derived contextual meaning of each item, the inner circle
the actual situational elements to which the forms relate.
7.5 We have said that translation equivalence occurs when SL
and TL items are relatable to ‘the same’ features of substance.
We will show how this applies in restricted translation in Chapters
8 and 9. Meanwhile we must refer to the problems raised by the
use here of the term ‘the same’.
7.51 In total translation, the question of ‘sameness’ of situation-
substance is a difficult one, and is linked to the question of the
‘sameness’ or otherwise of the cultures (in the widest and loosest
sense) to which SL and TL belong. ‘Situation’ in relation to
contextual meaning, is a wide blanket-term which, within a
general semetic theory, requires considerable refinement. Any
speech-act takes place in a specific bio-socio-physical environ-
ment, at a specific time and place, between specific participants
and so on. But the text which is (for the linguist) the central item
in the speech-act is, or may be, relatable not only to features of
this immediate situation, but also to features at greater and greater
distances (so to speak) reaching out, ultimately, into the total
cultural background of the situation. The ‘situation’, in other
words, may be thought of as a series of concentric circles, or
spheres, of relevance to the text. Something is said about the
relevance of wider or more peripheral situational features in the
chapters on Language Varieties in Translation and The Limits of
Translatability (13 and 14 below). Clearly, since translation
equivalence demands that SL and TL text should be relatable
to ‘the same features of substance’ there must be community of
relevant substance for the two texts.
7.6 Leaving aside the question of total translation, we see that
this necessity of ‘community of relevant substance’ for translation
enables us at once to posit the limits of translatability for restricted
52
CONDITIONS OF TRANSLATION EQUIVALENCE
translation. These limits are summed up in two generaliza-
tions :
(i) Translation between media is impossible (i.e. one cannot ‘translate’
from the spoken to the written form of a text or vice-versa) .
(ii) Translation between either of the medium-levels ( phonology and
graphology) and the levels of grammar and lexis is impossible (i.e. one
cannot ‘translate’ from SL phonology to TL grammar , or from SL
lexis to TL graphology . . . etc.) .
7.61 These generalizations may require a little elucidation.
As to (i) : the substance which is relevant to phonology is phonic
substance, and the substance which is relevant to graphology is
graphic substance. The substantial features relevant to a phono-
logical unit or item are sounds produced in a human vocal tract.
The substantial features relevant to a graphological unit or item
are visible marks on paper, stone . . . etc. Phonic and graphic
substance are absolutely different ; therefore there can be no question
of a phonological item being relatable to the same substantial
features as a graphological item.
For any particular language, of course, there is an arbitrary
relationship between graphological and phonological 3 units. Con-
version from spoken to written medium, or vice-versa, is a
universal practice among literates; but it is not translation, since
it is not replacement by items which are equivalent because of
relationship to the same substance.
7.62 For (ii), the same applies: phonic and graphic substance
are absolutely different from situation-substance. Translations occur
in which it looks, at first sight, as if a phonological item is being
translated by a grammatical item or items: e.g. when English:
You're going to Helsinki? with a final-rise intonation is translated
into Finnish as Menetle Helsinkiinko ? (with falling intonation, but
‘interrogative particle’ kb). Here it may seem as if an English tone
(a phonological item) has a Finnish grammatical item as its trans-
lation equivalent. But this is not so. In the SL text, the particular
tone is a phonological exponent of the grammatical category
‘enquiry’ and this in turn has a contextual meaning which
relates it to a feature of jttaa/ton-substance. This same feature of
• Or formal units, in the case of what is sometimes called logographic writing:
e.g. Chinese, Egyptian hieroglyphs.
53
A LINGUISTIC THEORY OF TRANSLATION
situation-substance is relatable to the Finnish grammatical cate-
gory of ‘interrogative’, whose exponent is kojko. There is com-
munity of substance between SL and TL; but the substance is
situation-snhs.ia.nce for both.
Another — negative — illustration of this is provided by an
Anglo-Yiddish joke, from which this is a relevant extract. 4
A Jew has been accused of horse-stealing, and, in court, the
following exchanges take place:
Judge : Did you steal a horse ?
Interpreter: Hot ir gestolen a pferd?
Accused: Ikh hob gestolen a A pferd?
Judge: What did he say?
Interpreter: He said ‘I stole a horse’.
The point here is that the interpreter has failed to take note of
a feature of the accused’s speech: namely, the Yiddish rise-fall
tone, A , which occurs on the word pferd. Had he done so, his
translation might have been ‘What. Me steal a horse?’ Now,
this looks, at first sight, like failure to translate a phonological
feature (the rise-fall tone) by a grammatical feature (a particular
class of clause). But the rise-fall tone is, of course, simply the
phonological exponent of a grammatical category of ‘incredulity’
— and this is relatable to the same situation-substance as is the
English clause-class SP, where the exponent of P is an infinitive,
or base-form, verbal group. 5
7.63 There are many other obvious cases where it looks as if a
phonological feature of an SL text has a lexical or grammatical
equivalent in the TL text. For example, in translation between
English and French it may happen that an ‘emphatic’ ‘I did it.’
is translated into French as ‘C’est bien moi qui l’ai fait.’ Now,
it looks, here, as if an English phonological feature — ‘marked
tonicity’, i.e. a special, contrastive, location of the tonic or tone-
bearing foot — is translated ‘grammatically’ by a special kind of
sentence- & clause-structure in French; similarly, ‘marked
* For this example I am indebted to my wife.
6 On the Yiddish rise-fall tone, cf. Uriel Wcinreich, ‘On the Yiddish Rise-
Fall Intonation Contour’, in For Roman Jukobson, p. 633.
54
CONDITIONS OF TRANSLATION EQUIVALENCE
tonicity’ may appear to be translated into Yukagir by special
‘assertive’ conjugations, marking the Subject, the Predicator
or the Complement as major-information-point 6 ; again, vowel-
lengthening in certain Javanese adjectives, translated into English
by the use of a (grammatical) ‘intensifying sub-modifier’, e.g.
terribly , very . . . etc.
In all these cases, the phonological feature (English tonicity,
Javanese vowel-lengthening) is merely the exponent of a grammatical
category; it is this grammatical category (not its phonological
exponent) which has a grammatical equivalent in the TL.
7.64 There is never any translation from phonology to gram-
mar ; nor from graphology to grammar. Thus in the first example
in the last section ‘7 did it’ is spoken of as if it represents a certain
phonological event. But looking at it as a piece of graphology in its
own right, we might have suggested that the graphological
feature italicization of I was translated grammatically into French.
This, of course, is not so ; italicization and tonicity are simply the
corresponding written and spoken exponents of the same gram-
matical category of major-information-point.
Translation from the levels of grammar and/or lexis to the
medium-levels is, of course, equally impossible. If someone says
of a given grammatical or lexical item ‘this can be translated into
English only by the tone of the voice ’ or something of that sort, this
must not be taken to mean that an English phonological feature
(tone or tonicity) is the translation equivalent of an SL gram-
matical/lexical feature. It means, simply, that the exponent of
the equivalent grammatical or lexical feature in English happens
to be tone or tonicity.
* cf. E. A. Kreinovich: Tukagirskij Yazyk (Moscow-Leningrad, 1958), pp.
131-38.
55
8
Phonological Translation
8.1 Phonological translation is restricted translation in which
the SL phonology of a text is replaced by equivalent TL phono-
logy. The grammar and lexis of the SL text remain unchanged,
except insofar as random grammatical or lexical deviations are
entailed in the process. Thus, as mentioned in 2.321 above, the
phonological translation of the English plural ‘cats’ / kats / into
a language which has no final consonant clusters might be some-
thing like / kat /. The phonological translation equivalent here
ends in / 1 / and thus appears to be a singular.
8. 1 1 The basis for translation equivalence in phonological
translation is relationship of SL and TL phonological units to
‘the same’ phonic substance. For example, the' phonological
translation of English ‘had’ / had / into Greek is / xent /. The
distinctive phonic substance related to English / h / is ‘voiceless
glottal fricative’ — i.e. a ‘deep’ voiceless fricative, that is, one in
which the fricative hiss is generated by turbulent air-flow through
the glottis and modulated by the vocoid-shaping of the mouth.
Greek has only one phoneme related to nearly the same phonic
substance, / x / — i.e. a ‘deep’ voiceless fricative, the hiss here
being generated by turbulent air-flow through a channel formed
between the dorsal surface of the tongue and the roof of the
mouth, and modulated by the vocoid-shaping of the mouth.
The English / a / is a low front vocoid, and the same phonic
features are present in Gk. / e / (although, in fact, the Gk. vowel
is not so low as the English one — but each is the lowest in the
front series of each language). English / d / is a voiced apical
stop. Gk. has an apical stop / 1 /, but in Greek the components
‘stop’ and ‘voice’ co-occur only when a nasal precedes. The
translation equivalent of English / d / therefore must either be
Greek / 1 /, or Greek / nt / manifested phonetically as [nd] :
Greeks normally use the latter when speaking English with a
‘Greek accent’ i.e. in phonological translation.
56
PHONOLOGICAL TRANSLATION
We may roughly tabulate these relationships in a diagram
analogous to that used in 5.43.
English
h
a
d
Greek
voiceless
deep (glottal)
,, (dorsal)
vocoid-shaped
fricative
lowest
front
vocoid
apical
voiced
(nasal +)
stop
Because Greek has only one vowel in the low front region /e/
as opposed to the two /e, a/ of English, and because of the relation
between voicing of stops and a preceding nasal, Greek has a
single phonological translation-equivalent, /xent/, in phonetic
transcription (xfindj, for the three English phonological forms
/hed/ /had/ /hand/. 1
8.2 In phonological translation, as in translation at other levels,
one must distinguish between formal correspondence and trans-
lation equivalence. We may take as an example a small sub-
system of phonemes, labial stops in English and Sindhi. The
English system is one of two terms / p / and / b /; the Sindhi
system is one of five terms / p / / ph / / b / / bh / / 6 /. Formally,
then, there can be no correspondence between the English and
Sindhi terms. It is possible, however, to set up translation
equivalences by considering the features of phonic substance to
which the English and Sindhi phonological units are related.
In terms of distinctive phonic features the Sindhi system is
1 A fact which is very troublesome for Greek beginners in English, the
situation being further complicated, as D. Abercrombie has pointed out
(Problems and Principles, Longmans Green, 1956 pp. 11 and 26) by the fact
that the modern Greek is a lexical translation-equivalent of English arm
as well as hand.
57
A LINGUISTIC THEORY OF TRANSLATION
3-dimensional; each term is either unaspirated (p b B) or aspirated
(ph bh), voiceless (p ph) or voiced (b bh 6), pulmonic egressive
(p ph b bh) or glottalic ingressive (6). The system is complicated
by syncretisms: (6) is not only glottalic ingressive, but also
voiced and unaspirated. The English system is unidimensional,
again slightly skewed by syncretisms; / p / is voiceless (and usually
aspirated) , / b / is unaspirated (and commonly voiced) . The
following system-diagrams roughly indicate the formal differ-
ences and translation equivalences.
Sindhi English
I
voiceless I voiced
The dotted outline of the English system-diagram indicates the
limits, in phonic substance, of the Sindhi system — limits rarely,
if ever, reached by the English phonemes; on the other hand,
English / b / is shown as extending into the phonic-substance
areas of Sindhi / p / / 6 / and / bh /. In general, superimposition
of the two diagrams roughly indicates both the non-correspond-
ence of the phonological systems, and the possibilities of trans-
lation equivalence.
8.21 The normal Sindhi translation equivalent of English / b /
is Sindhi / b / since both are related to voice and unaspirated as
features of phonic substance. However, English / b / can, on
occasion, be voiceless unaspirated, or realized as a weak glottalic
ingressive, or followed by a ‘whispery-voiced’ vowel (i.e. be ‘voiced
aspirated’); there may thus be occasions when the Sindhi trans-
lation equivalent of Eng. / b / may be / p /, / 6 / or / bh /. In
58
PHONOLOGICAL TRANSLATION
other words, we can say ‘the highest probability Sindhi trans-
lation equivalent of Eng. / b / is / b /; but in special co-textual
conditions other equivalents may occur’. And, by taking note of
the special factors involved we could, in theory, determine the
conditioned probabilities of these equivalences.
8.3 Phonological translation is thus seen to parallel ‘total
translation’ quite closely; for any one SL phonological item there
may be more than one TL phonological translation equivalent.
The particular TL equivalent depends on what particular
features of phonic substance are relatable to the SL item on that
particular occasion; precisely as the particular English trans-
lation equivalent of Navaho dootl’ii depends on what specific
feature of situation-substance (what colour) is relatable to the
Navaho item on that occasion.
8.4 Phonological translation, like total translation, may involve
change of rank, or regrouping and reorganizing of features of
substance into the formal units of the TL. For example, in
phonological translation between English and Japanese, equi-
valence often has to be established not merely at phoneme rank,
but with an upward change of rank, to the next higher unit in
the Japanese phonological hierarchy — namely the mora or ‘kana’
(somewhat similar to, though neither formally nor substantially
identical with the English syllable). Thus English platonic love ,
phonologically (ignoring, for the present, anything above the
phoneme rank in English) / pl a tonik lav / has as its Japanese
phonological translation equivalent / puratonikkurabu /.
8.41 We may show the translation equivalences here as
follows :
English p l a to ni k la v
Japanese pu ra to ni k-ku ra bu
The English phonemes / p / and / v / have, as Japanese trans-
lation equivalents, the morae / pu / and / bu /; and the phoneme
I k / has, as its translation equivalent, two Japanese morae
/ k-ku /. The reason for these equivalences is that, except in
certain definable circumstances, the Japanese mora always has
the structure CV; an English C before C, or finally, is thus
normally represented by a Japanese CV structure. Moreover,
59
A LINGUISTIC THEORY OF TRANSLATION
the Japanese high vowels / u / and / i / are often realized voiceless
or as phonic zero-, hence / pu-ra / is most nearly relatable to the
same phonic substance as English / pi /, with the ‘aspiration’ of
/ p / manifested phonetically as a partially voiceless / I /.
The equivalence / k / = / k-k u / requires further comment.
Here the Japanese translation treats, as phonologically relevant,
a feature which is present in the phonic substance of the English
utterance — namely, length of / k / after a short vowel — but is not
phonologically relevant in English. This is precisely paralleled in
total translation by, for example, Russian treating a situational
feature (on foot) as linguistically relevant in pnUa for English
have arrived (see 5.4 above).
8.5 Another example of the reorganization of phonic substance
into TL phonological units which may occur in phonological
translation is provided by the following true story 2 .
A Scotsman in France went to buy an ice-cream cone at a
kiosk. Two types were available — one with a single ball of ice-
cream, the other with two balls of ice-cream side by side. The
Scotsman wanted one of these double helpings, so he asked in
English, for ‘a double’. He was unhesitatingly served with the
type required.
8.51 Assuming that the ice-cream vendor was monolingual
(which is almost certain) what happened was probably this: the
English / dabl / (with a ‘dark’, or strongly velarized / I / ) was
‘translated’ at the phonological level into / da bul /, and inter-
preted by the vendor as ‘deux boules’ — the correct French
technical term for the type of ice-cream required.
The similarity of phonic substance justifying the translation
equivalent of French / d /, / b / to English / d /, / b / requires
no comment. The equivalence / s / = / 0 / is easily explained;
the phonic substance: ‘obscure (i.e. not clearly front-unrounded
or back rounded) or “mixed” vocoid’ is common to / a / and
/ 0 /. The explanation of / I / = / ul / is more interesting. Here
the phonic features ‘laterality’ and 'velaritf, simultaneously pre-
sent in the phonic substance of English / 1 / are redistributed in
translation into a sequence of two phonemes : / u / (incorporating
velarity) and / I / (incorporating laterality).
2 I am indebted to my colleague J. McH. Sinclair for this example.
60
PHONOLOGICAL TRANSLATION
A parallel in total translation would again be the translation
of Russian pnsla by arrived on foot ; here a feature of situation
substance which is related to the single lexical item PRIITI in
Russian must, if incorporated into the English translation at all,
be redistributed as indicated.
8.6 Phonological translation, as we remarked in 2.323, is
practised deliberately by actors and mimics when they assume a
foreign, or dialectal, pronunciation. It can occur, receptively, as
in the example just given, and it occurs productively, though
inadvertently, in the imperfect pronunciation of someone speak-
ing a foreign language. We usually call this latter ‘phonological
(or phonetic) interference’ and think of it as a transference of
native (Lj) phonology into the foreign language (L 2 ) being
spoken. This is a perfectly possible way of describing the pheno-
menon. From the point of view of translation-theory, however,
it may be regarded as translation from the L 2 into the L 1; at the
phonological level only, since it is replacement of L 2 phonology
by equivalent L, phonology (lexis and grammar remaining
unchanged).
In normal total translation the SL phonology is not translated,
but merely replaced by whatever (non-equivalent) TL phonology
is entailed by the selection of TL grammatical and lexical items.
In certain circumstances, however, the translator attempts to
reproduce at least some features of SL phonology in the TL text —
i.e. performs a partial phonological translation, and this, in turn,
affects the grammatical/lexical translation, since the selection of
translation equivalents at these formal levels is partly determined
by the need for their phonological exponents to be translation
equivalents of phonological items in the SL. This happens
typically in Jilm-dubbing, where the translator may select lexical
translation equivalents in the TL which have labials, for instance,
in their phonological forms, to match labials in the phonological
forms of the SL items. In the translation of poetry, too, some
attempt may be made to select TL equivalents which ‘match the
sound’ of SL items; this entails some degree of phonological
translation.
61
9
Graphological Translation
9. 1 Graphological translation is restricted translation in which
the SL graphology of a text is replaced by equivalent TL
graphology. The basis for equivalence is relationship to ‘the same’
graphic substance.
The discussion of graphological translation is more difficult
than the discussion of phonological translation because we have
no systematic theory of graphic substance — no ‘general graph-
etics’ — from which to draw categories for the description of
graphic substance. We therefore construct ‘graphetic categories’
ad hoc for the present section.
9.2 We can illustrate graphological translation fairly simply by
means of the Cyrillic and Roman alphabets. To simplify the
discussion still further, we will confine ourselves to capitals.
9.21 On the basis of a rough graphetic analysis, we can say
that the graphological units, or letters, of the two alphabets make
use of eight distinctive features of graphic substance, subject to
certain systematic modifications. These are: vertical | , horizontal
— , right oblique / , right oblique curved , left oblique \ , right semi-
circle 3 , left semicircle C , supine semicircle W.
The modifications are: full 1 / — D etc., low i/_o etc., high
etc.and(for only) mid — . Full can be unmarked: other
modifications indicated by l, h, m. (An additional modification,
reduced (r) is occasionally useful.)
Finally, there are categories of combination, e.g. (a) attached,
(x) crossing, (c) connecting. If graphic features are named in a
sequence corresponding to left-to-right, and upwards for ascenders
(verticals and obliques), attachment points can usually be left
undescribed, and normal end-attachment can be assumed except
when otherwise stated.
Thus, in both Cyrillic and Roman we have, for instance:
A = / \ — me B = | 3 h D 1
H = 1 | — me T = | — hx
X=/\x K=J/h\l
62
GRAPHOLOGICAL TRANSLATION
For Cyrillic only:
5 = ! — h D l r = 1 — h
H = I 1/ C fi = ) — hrl
= J — hr i — 1 I Ira i Ira M = 3 | C — me
[° r * = ( | / \ ) x]
9.22 The appropriate graphological translation equivalents are,
in many cases, obvious. The following Cyrillic letters, however,
present problems:
BrA?ie3yi ny#Mmuai,*bbi baiofl
9.23 We shall analyse a few in detail.
5 : distinctive features: | — h 3 1: i-e., counting modifications,
5 features.
Several Roman letters have 2 or 3 features in common:
D ( I 3 ), T ( [ - hx), F ( | — h -m), B ( 1 3 h 3 1).
The attachment point of the high horizontal in T (‘crossing’)
tends to rule T out. The choice lies between F and B, both
of which have 3 features in common with 5 • We decide,
somewhat arbitrarily or impressionistically, for B.
p | — h Possible translation-equivalents: L ( | — 1),
F ( |< — h — m), T ( | — hx). F has all three features of f~,
L has only two. However, F introduces two additional fea-
tures ( — and m). We therefore choose L.
A : J I — hre — lc | Ira | Ira No close equivalent. Best is
possibly A, which, like A> has two ascenders, and a hori-
zontal connector. But A is almost untranslatable.
)j( : 3 | C 111X0 close equivalent. Redistribution into
two letters is almost essential. DC is possible, this has 3 fea-
tures in common ( | 3 C )» though rearranged. HC
(|| — me, Q ) has 4 common features. We select HC.
'3> : D h D 1 Only possible equivalent: S ( C h 3 1)
n : J — hre | No single letter equivalent. Best approximation
isJT | , | - hx)
63
A LINGUISTIC THEORY OF TRANSLATION
9.24 We summarize the rest:
FI ( I — he | ) = H ( | _ me | ),
y (\hy ) =y(\ h / h i i),
$ ( C 31 *) = Q. ( C 3 I xlr )>
U ( I — lc 1 ) = U ( I U lc | ),
M(^h ) ) =J (ul | ),
LLI ( [ I I — lc) — no near equivalent; use W ( \/ 1 \ 1 / ),
LU.( 1 l 1 — lc) l lr)— useW, 3 ( D - m) = D ( D l ),
K>( I - mrc C 3 ) = 10, * ( / 1 C h I ) = R ( I 3 h \ 1).
"b 1 and b are difficult. No Roman letter has ‘low reduced right
semicircle’. Ignoring this we can translate b I as II, b as I (if
lower-case were used, the equivalents would be bl and 1). '"b can
only be T or Z. Neither is very close: we choose Z.
9.3 Here, again, in graphological translation, we see that
phenomena analogous to those of total translation occur. First,
the TL equivalent is seldom related to exactly the same set of
substance-features as the SL item. Secondly a single unit of SL
may have to be redistributed into two TL units.
9.31 The following is an example of graphological translation
from Russian (Cyrillic) into Roman caps.
DfcEJlESHA* ^OPOTA = HCEJTESHAR AOPOLA
enyTH H K = CHYTHNK
9.32 Above we gave a graphological translation of sputnik from
Russian italic, or cursive, (cnymmiK) as cnymhuk. This, again,
illustrates a phenomenon found in total translation. The differ-
ence between caps, and lower-case is one of medium-variety, and
is thus more or less equivalent to a difference at the grammatical/
lexical levels between dialects, registers or styles. Different vari-
eties, of course, have different characteristics and call for different
TL equivalents.
9.4 Graphological translation is often particularly difficult
because writing systems tend to make use of a restricted range
of graphic substance; the straight lines and sharp angles of Ar-
menian script, for instance, have little substance in common with
the curves and circles of Burmese, or some South Indian scripts.
64
GRAPHOLOGICAL TRANSLATION
Occasionally one can make some degree of graphological
translation between mutually exotic scripts, but this is often
limited in extent. The following, for instance, is a graphological
translation from a cursive Roman into Arabic:
J-Lrabic
The translation does some slight violence to Arabic writing
conventions, but apart from that, the translation equivalences
can all be justified by relation of English and Arabic letters to
similar features of graphic substance.
9.5 An approximation to graphological translation is occasion-
ally practised deliberately by typographers who wish to give an
‘exotic’ flavour to written texts. For example, books about Islam
or the Arab world sometimes have their titles written in somewhat
Arabic-looking script — a graphological semi-translation..
Persons writing in a foreign language may occasionally pro-
duce graphological translations; for example, Greeks writing
in English (or in Roman in general) often replace a script
a by a, or an n by 7). It is particularly clear in the case of 7)
for n that this is graphological translation, since the only
thing in common between n and 7) is relation to similar graphic
substance.
65
Transliteration
10.1 It will now be clear that graphological translation is quite
different from transliteration. We repeat here an example from
9.31 with a transliteration added.
Original: GnyTHMK
Graphological Translation: CHYTHNK
Transliteration: SPUTNIK
In transliteration, SL graphological units are replaced by TL
graphological units; but these are not translation equivalents,
since they are not selected on the basis of relationship to the same
graphic substance.
In the process of actually transliterating a text, the trans-
literator replaces each SL letter or other graphological unit by a
TL letter, or other unit, on the basis of a conventionally estab-
lished set of rules. The transliteration rules specify transliteration-
equivalents which differ from translation equivalents in two ways :
first, in not necessarily being relatable to the same graphic
substance as the SL letters; secondly, in being (in good trans-
literation) in one-to-one correspondence with SL letters or other
units.
10.2 In principle, the process of setting up a transliteration-
system involves three steps:
(i) SL letters are replaced by SL phonological units; this is the
normal literate process of converting from the written to the
spoken medium.
(ii) The SL phonological units are translated into TL phono-
logical units.
(iii) The TL phonological units are converted into TL letters,
or other graphological units.
In a simplified way this process may be indicated as follows.
Transliteration here is from Russian into English.
66
TRANSLITERATION
SL‘ graph.
SL phon.
TL phon.
TL graph.
units
units =
units
units
1 . 5
lb/ =
N
B
2. B
M =
M
V
3. 4
A// =
A//
CH
4. X
M -
N
H etc.
This table is to be interpreted thus : in 1 . the Russian (Cyrillic)
graphological unit 5 is convertible into the Russian phonological
unit jbj. This / b/ has phonic features (voice, labiality, stopness)
in common with the English phonological unit / b/, hence Eng-
lish / b/ is its English phonological translation-equivalent. The
English phonological Unit /b/ is convertible into the English
graphological unit B. The letter B is thus the English translitera-
tion-equivalent of the Russian 6. In 3. we have a single Russian
letter 4 with its phonological counterpart jtjj. The English
phonological translation-equivalent is /t f/, but this has no single
letter exponent in English graphology — it is convertible to CH.
Hence the English letter-sequence CH is the transliteration-
equivalent of the Russian single letter HL .
10.3 There are, however, several corr dicating factors.
(i) A given SL letter may have more than one SL phonological
correspondent. In this case only one of these must be chosen as
basis for transliteration: e.g. in transliteration from English into
Russian we could find that Eng. C -> either / k / or is j. For
Russian transliteration equivalent we should have to choose
between the Cyrillic letters K and C.
(ii) In phonological translation between SL and TL there may
not be — indeed normally is not — one-to-one equivalence between
SL and TL phonological units; where two or more SL units have
the same TL phonological translation equivalent an arbitrary
distinction must be introduced into the TL graphological repre-
sentation of these units. For example, in transliteration from
Sanskrit (Devanagari) into Roman (say, English) we would find
the following situation :
SL graph. SL phon. TL phon. TL graph,
units units unit unit
SH
SH, arbitrarily split — ►
SH
51
K|
0
= /(7
67
A LINGUISTIC THEORY OF TRANSLATION
(iii) In converting from TL phonological to graphological units
there may be a selection of letters to choose from; a decision
must be made arbitrarily.
Thus, in transliterating, say, Russian K into English we can
set up the phonological translation-equivalent /k/ = /k/. We
then have to choose between converting /k / -> K or -*■ G. We
might make an arbitrary choice, say C. Or we might decide to
transliterate K by K, thus releasing C as the transliteration-
equivalent of .
10.4 A particular complication arises when the TL grapho-
logical units are not immediately convertible to TL phonological
units. This happens in languages where the writing system is
‘logographic’ 1 , e.g. Chinese. The Chinese graphological unit —
the character — is directly convertible to a lexical or grammatical
unit of the language. ‘Transliteration’ in such a case can only
be through the phonological form of the lexica] or grammatical
unit associated with the character.
In other words, given the character /v we can ‘transliterate’
into English only by first converting the character J\r into the
lexical item which it represents, and then ‘transliterating’ that as,
say, ren. But this is not transliteration, in that the graphological
units of the TL form are not in one-to-one correlation with
graphological units of the SL. It is, on the contrary, a form of
transcription (see 10.6 below).
10.5 One final complication must be mentioned. Translitera-
tion is a conventionalized process, unlike translation which is
carried out anew, or ad hoc, on each particular occasion. In
transliteration, what we have so far called the ‘TL’ may not,
strictly, be a ‘language’ (or the writing system of a specific
language) at all.
Thus, for example, the traditional transliteration of Sanskrit is
into Roman letters — but it is not, strictly, into Latin graphology,
for the Roman alphabet has to be supplemented by a number of
diacritics to correspond to letter-distinctions in Skt. which are not
present in Latin.
10.6 In 10.4 above we mentioned transcription. A transcription is
1 Often erroneously called ‘ideographic’ or ‘pictographic’.
68
TRANSLITERATION
a writing-system in which the letters or graphological units are in
one-to-one correlation with phonological units — or with spans or
segments of phonic substance. In the first case we have a phono-
logical transcription (which may be phonemic, allophonic, prosodic,
etc.) and in the second, a phonetic transcription.
The distinction between transliteration and transcription is
important, and often misunderstood. For example, it has been
argued that Hebrew cannotbe ‘Romanized’ becauseRomanization
distorts the language; i.e. fails to show the formal and phono-
logical relatedness of lexically or grammatically related forms.
The source of this contention is the fact that it is tacitly
assumed that a Romanized writing system for Hebrew will
necessarily be a phonemic transcription. Now, a phonemic
transcription of Hebrew may, indeed, do precisely what is feared;
to understand why, one must remember that the Hebrew script,
like the Arabic script, is basically syllabic. Each letter represents,
essentially, a specific consonant, with the implication of a follow-
ing (unspecified) vowel. The particular vowel can, if desired, be
indicated by a diacritic mark; in addition, suppression of the
implied vowel can also be marked. Moreover, certain letters
represent (syllables beginning with) consonants which alternate
between a stop and a. fricative stricture-type.
In traditional Hebrew orthography the ‘consonantal shape’ of
a lexical item is preserved throughout its paradigmatic morpho-
logical changes, because the varying vowels, and the consonantal
alternations, are either not graphically represented, or repre-
sented only by diacritics; thus, for the verb ‘to write’ we have
the following forms :
Infin.
Imper.
3rd sg. past
present etc.
ti
tins
ins
hlVD
In transcription:
lixtov ktov katav kot§v etc.
But, in true transliteration:
Iktb ktb ktb kwtb
A LINGUISTIC THEORY OF TRANSLATION
or, with diacritic vowels:
hkt°b kt°b k a t a b k^wt^b
Transcription is a representation of phonological units: trans-
literation , however, gives a one-to-one representation of grapho-
logical units, and consequently can represent precisely the
traditional Hebrew orthography, and preserve the visual related-
ness between forms, which a phonemic transcription tends to
obscure.
70
Grammatical and Lexical Translation
11.1 Grammatical translation is restricted translation in which the
SL grammar of a text is replaced by equivalent TL grammar,
but with no replacement of lexis. The basis for equivalence here,
as in total translation, is relationship to the same situation-
substance.
11.11 Thus, given an English SL text such as This is the man 1
saw, we might translate it grammatically into French as Void le
man que j'ai see-e: or into Arabic as haada 'l-man ’ili see-t-u. In both
of these translations we have retained the two lexical items, man
and see, unchanged, but have replaced all the grammatical items
by equivalent French or Arabic grammatical items.
11.12 In more detail, the process of grammatical translation in
the Arabic example is as follows :
English clause-structure SPC = Arabic SPC or SC; the latter
being translation equivalent of an English SPC structure in
which P — be (present tense), as here. So here we have SPC = SC.
The exponent of S in the English text is the item this , a term
in the system of English deictics; the Arabic translation equi-
valent is haada. The exponent of C in English is the Ngp the man
l saw, i.e. a Ngp with the structure MHQin which the exponent
of M is the definite article the for which the Arabic translation
equivalent is al. The exponent of H is the lexical item man, which
remains unchanged. The exponent of Q is a rank-shifted clause
of structure SP. The Arabic equivalent is a rank-shifted clause of
structure C A S P C p e _ w ith connective C A (’ili) and a complex
predicator with bound subject-object morphemes: (an approxi-
mate morpheme-rank-bound total back-translation of the Arabic
C would be the man which see-d-l-him ) .
Hence the grammatical translation:
Haada al-man ’ ili see-t-u.
1 1.2 Lexical translation is restricted translation in which the SL
lexis of a text is replaced by equivalent TL lexis, but with no
71
A LINGUISTIC THEORY OF TRANSLATION
replacement of grammar. The basis of equivalence again is
relationship to the same situation-substance.
11.21 Thus our same text, This is the man I saw , translated
lexically into French and Arabic would be: Fr. This is the homme
I voi-ed and This is the rajul I shuf-ed.
11.22 Here, the English SL grammar is preserved, but the
exical items man and see are replaced by the equivalent TL
items hommejrajul and voi-jshuf.
1 1 .23 Now it is at once evident that, unlike grammatical
translation, this process, or one very like it, occurs in real life.
British soldiers in the Middle East have often produced utterances
not unlike ‘This is the rajul I shufed’. In other words, the process
of ‘picking up a few words’ of the language, and then throwing
them into utterances in the speaker’s primary language involves
lexical translation — rarely, if ever, grammatical translation.
1 1 .3 Since grammar and lexis between them exhaust the
formal levels of language, grammatical and lexical translation
between any two languages are the converse of each other; that
is to say, grammatical translation from language A into language
B is the same as lexical translation from language B into language A.
The following diagram shows the relations between gram-
matical, lexical, and total translation:
Haada al-man ’ Hi see-t-u
This is the rqjul I shufed
72
12
Translation Shifts
Having reviewed all types of restricted translation we return,
now, to general discussion; in particular, to a brief systematic
survey of some of the changes or ‘shifts’ which occur in transla-
tion. By ‘shifts’ we mean departures from formal correspondence
in the process of going from the SL to the TL. Two major types
of ‘shift’ occur: level shifts (12.1} and category shifts (12.2).
12.1 Level shifts. By a shift of level we mean that a SL item at
one linguistic level has a TL translation equivalent at a different
level.
We have already pointed out (7.6) that translation between
the levels of phonology and graphology — or between either of these
levels and the levels of grammar and lexis— is impossible. Trans-
lation between these levels is absolutely ruled out by our theory,
which posits ‘relationship to the same substance’ as the necessary
condition of translation equivalence. We are left, then, with
shifts from grammar to lexis and vice-versa as the only possible
level-shifts in translation; and such shifts are, of course, quite
common.
12.1 1 Examples of level shifts are sometimes encountered in the
translation of the verbal aspects of Russian and English. Both
these languages have an aspectual opposition — of very roughly
the same type — seen most clearly in the ‘past’ or preterite tense:
the opposition between Russian imperfective and perfective (e.g.
pisal and napisal) , and between English simple and continuous ( wrote
and was writing).
There is, however, an important difference between the two
aspect systems, namely that the polarity of marking is not the same.
In Russian, the (contextually) marked term in the system is the
perfective ; this explicitly refers to the uniqueness or completion of the
event. The imperfective is unmarked — in other words it is relatively
neutral in these respects (the event may or may not actually be
unique or completed, etc., but at any rate the imperfective is
73
A LINGUISTIC THEORY OF TRANSLATION
indifferent to these features — does not explicitly refer to this
‘perfectiveness’) . 1
In English, the (contextually and morphologically) marked
term is the continuous; this explicitly refers to the development,
the progress , of the event. The ‘simple’ form is neutral in this
respect (the event may or may not actually be in progress, but
the simple form does not explicitly refer to this aspect of the
event) .
We indicate these differences in the following diagram, in
which the marked terms in the Russian and English aspect-
systems are enclosed in rectangles:
in progress
Event
repeated
unique, completed
p i s a 1
n a p i s a 1
was writing
wrote
12.12 One result of this difference between Russian and English
is that Russian imperfective (e.g. pisal) is translatable with almost
equal frequency by English simple (wrote) or continuous (was
writing). But the marked terms (napisal — was writing) are
mutually untranslatable.
A Russian writer can create a certain contrastive effect by
using an imperfective and then, so to speak, ‘capping’ this by
using the (marked) perfective. In such a case, the same effect of
explicit, contrastive, reference to completion may have to be
translated into English by a change of lexical item. The following
example 2 shows this:
1 My attention was first drawn to this difference between English and
Russian by Roman Jakobson in a lecture which he gave in London in 1950.
2 From Herzen, cited by Unbegaun in Grammaire Russe, p. 217.
74
TRANSLATION SHIFTS
‘Cto it delal Bcl'tov v prodolzenie etix des'ati let? Vse
il pocti vse. Cto on sdelal ? Nicego ili pocti nicego.’
Here the imperfective, delal, is ‘capped’ by the perfective
sdelal. delal can be translated by either did or was doing — but, since
there is no contextual reason to make explicit reference to the
progress of the event, the former is the better translation. We can
thus say ‘What did Beltov do . . . ?’ The Russian perfective,
with its marked insistence on completion can cap this effectively:
‘What did he do and complete ?’ But the English marked term
insists on the progress of the event, so cannot be used here. (‘What
was he doing’ is obviously inappropriate.) In English, in this case,
we must use a different lexical verb : a lexical item which includes
reference to completion in its contextual meaning, e.g. achieve s .
The whole passage can thus be translated :
‘What did Beltov do during these ten years? Everything,
or almost everything. What did he achieve? Nothing, or
almost nothing.’
12.13 Cases of more or less incomplete shift from grammar to
lexis are quite frequent in translation between other languages.
For example, the English : This text is intended for . . . may have
as its French TL equivalent: Le present Manuel s' adresse a. .. .
Here the SL modifier, This — a term in a grammatical system of
deictics — has as its TL equivalent the modifier Le present, an
article -f- a lexical adjective. Such cases are not rare in French,
cf. also This may reach you before I arrive — Fr. II se peut que ce mot
vous parvienne avant mon arrivee. Once again the grammatical item
this has a partially lexical translation equivalent ce mot.*
12.2 Category shifts. In 2.4 we referred to unbounded and rank-
bound translation : the first being approximately ‘normal’ or ‘free’
translation in which SL-TL equivalences are set up at whatever
rank is appropriate. Usually, but not always, there is sentence-
* Another possibility would be ‘What did he get done?’, but this would be
stylistically less satisfactory.
* Examples from Vinay et Darbelnet, Stylistique Comparie du fran(ais el de
I'anglais, p. 99.
75
A LINGUISTIC THEORY OF TRANSLATION
sentence equivalence®, but in the course of a text, equivalences
may shift up and down the rank-scale, often being established at
ranks lower than the sentence. We use the term ‘rank-bound’
translation only to refer to those special cases where equivalence
is deliberately limited to ranks below the sentence, thus leading to
‘bad translation’ = i.e. translation in which the TL text is either
not a normal TL form at all, or is not relatable to the same
situational substance, as the SL text.
In normal, unbounded, translation, then, translation equi-
valences may occur between sentences, clauses, groups, words
and (though rarely) morphemes. The following is an example
where equivalence can be established to some extent right down
to morpheme rank :
Fr. SL text J’ai laisse mes lunettes sur la table
Eng. TL text I’ve left my glasses on the table
Not infrequently, however, one cannot set up simple equal-
rank equivalence between SL and TL texts. An SL group may
have a TL clause as its translation equivalent, and so on.
Changes of rank (unit-shifts) are by no means the only changes
of this type which occur in translation ; there are also changes of
structure, changes of class, changes of term in systems, etc. Some of
these — particularly structure-changes — are even more frequent than
rank-changes.
It is changes of these types which we refer to as category-shifts.
The concept of ‘category-shift’ is necessary in the discussion of
translation; but it is clearly meaningless to talk about category-
shift unless we assume some degree of formal correspondence
between SL and TL ; indeed this is the main justification for the
recognition of formal correspondence in our theory (cf. Chapter
4). Category-shifts are departures from formal correspondence in
translation.
We give here a brief discussion and illustration of category-
shifts, in the order structure-shifts, class-shifts, unit-shifts (rank-
changes), intra-system-shifts.
* W. Freeman Twaddell has drawn my attention to the fact that in German-
English translation, equivalence may be rather frequently established between
the German sentence and an English unit greater than the sentence, e.g.
paragraph.
76
TRANSLATION SHIFTS
12.21 Structure-shifts. These are amongst the most frequent
category shifts at all ranks in translation; they occur in phono-
logical and graphological translation as well as in total translation.
12.211 In grammar , structure-shifts can occur at all ranks. The
following English-Gaelic instance is an example of clause-structure
shift.
SL text John loves Mary — SPC
TL text Tha gradh aig Iain air Main = PSCA
(A rank-bound word-word back-translation of the Gaelic TL
text gives us : Is love at John on Mary)
We can regard this as a structure-shift only on the assumption
that there is formal correspondence between English and Gaelic.
We must posit that the English elements of clause-structure
S, P, C, A have formal correspondents S, P, C, A in Gaelic ; this
assumption appears reasonable, and so entitles us to say that a
Gaelic PSCA structure as translation equivalent of English SPG
represents a structure-shift insofar as it contains different elements.
But the Gaelic clause not only contains different elements — it
also places two of these (S and P) in a different sequence. Now,
if the sequence SP were the only possible sequence in English (as
^ ...
PS is in Gaelic) we could ignore the sequence and, looking only at
the particular elements, S and P, say that the English and Gaelic
structures were the same as far as occurrence in them of S and P
was concerned. But sequence is relevant in English and we
therefore count it as a feature of the structure, and say that, in
this respect, too, structure-shift occurs in the translation.
12.212 Another pair of examples will make this point clearer
by contrasting a case where structure-shift occurs with one where
it does not.
I in the boat
A
A
j anns a' bhata
A. English
Gaelic
The man /
S
P
Tha
j an. duine
77
A LINGUISTIC THEORY OF TRANSLATION
and
B. English Is / the man / in the boat?
PS A
I I I
PS A
Gaelic Am bheil / an duine / anns a' bhata?
In B, there is complete formal correspondence of clause-
structure (no structure-shift) : in A, there is a structure-shift at
clause-rank.
These two examples, in fact, provide us with a commutation
which establishes the following translation equivalences:
A. English (SP) Gaelic VA a t P
B. English (SP) Gaelic V 1 at P
In other words, the Gaelic translation equivalent of the English
sequence -»• of S and P in clause-structure is the occurrence in
Gaelic of a verbal group of the class Affirmative as exponent of P;
the Gaelic translation equivalent of the English sequence ■*- of
S and P in clause-structure is the occurrence in Gaelic of a verbal
group of the class Interrogative as exponent of P.
These two examples in fact illustrate two different types of
translation-shift; in A, there is structure-shift; in B, there is unit-
shift, since in this case the Gaelic equivalent of a feature at
clause-rank is the selection of a particular term in a system
operating at group rank.
12.21 3 Structure-shifts can be found at other ranks, for example
at group rank. In translation between English and French, for
instance, there is often a shift from MH (modifier + head) to
(M) HQ, ((modifier +) head + qualifier), e.g. A white house (MH)
= Une maison blanche (MHQJ.
•12.22 Class-shifts. Following Halliday, we define a class as ‘that
grouping of members of a given unit which is defined by opera-
tion in the structure of the unit next above’. Class-shift, then,
occurs when the translation equivalent of a SL item is a member
of a different class from the original item. Because of the logical
dependence of class on structure (of the unit at the rank above)
it is clear that structure-shifts usually entail class-shifts, though
78
TRANSLATION SHIFTS
this may be demonstrable only at a secondary degree of delicacy.
For example, in the example given in 12.213 above (a white
house = une maison blanche), the translation equivalent of the
English adjective ‘white’ is the French adjective ‘blanche’. Insofar
as both ‘white’ and ‘blanche’ are exponents of the formally
corresponding class adjective there is apparently no class-shift.
However, at a further degree of delicacy we may recognize two
sub-classes of adjectives; those operating at M and those opera-
ting at Q in Ngp structure. (Q-adjectives are numerous in
French, very rare in English.) Since English ‘white’ is an M-
adjective and French ‘blanche’ is a Q-adjective it is clear that the
shift from M to Q entails a class-shift.
In other cases, also exemplified in the translation of Ngps from
English to French and vice-versa, class-shifts are more obvious:
e.g. Eng. a medical student = Fr. un etudiant en medecine. Here the
translation equivalent of the adjective medical, operating at M,
is the adverbial phrase en medecine, operating at Q_; and the
lexical equivalent of the adjective medical is the noun medecine.
12.23 Unit-shift. By unit-shift we mean changes of rank — that
is, departures from formal correspondence in which the trans-
lation equivalent of a unit at one rank in the SL is a unit at a
different rank in the TL.
We have" already seen several examples of unit shift in what
precedes: e.g. in sections 3.222, 3.223, 8.41, 12.211, 12.213. A
more appropriate term might be ‘rank-shift’, but since this has
been assigned a different, technical, meaning within Halliday’s
theory of grammar we cannot use it here.
12.24 Intra-system shift. In a listing of types of translation-shift,
such as we gave in 12.2 above, one might expect ‘system-shift’ to
occur along with the names of the types of shift affecting the other
fundamental categories of grammar — unit, structure and class.
There is a good reason for not naming one of our types of shift
‘system-shift’, since this could only mean a departure from formal
correspondence in which (a term operating in) one system in the
SL has as its translation equivalent (a term operating in) a
different — non-corresponding — system in the TL. Clearly, how-
ever, such shifts from one system to another are always entailed
by unit-shift or class-shift. For instance, in example B in 12.212
79
A LINGUISTIC THEORY OF TRANSLATION
the Gaelic equivalent of English clause-structure PS is shown to
be selection of a particular class of Verbal group (V^). We could
say that here there is a system-shift, since PS, a term in a system
of clause-classes, is replaced by V^, a term in a (formally non-
corresponding) system of Vgp classes. There is no need to do this,
however, since such a shift is already implied by the unit-shift.
We use the term intra-system shift for those cases where the
shift occurs internally, within a system; that is, for those cases
where SL and TL possess systems which approximately corres-
pond formally as to their constitution, but when translation
involves selection of a non-corresponding term in the TL system.
It may, for example, be said that English and French possess
formally corresponding systems of number. In each language, the
system operates in nominal groups , and is characterized by concord
between the exponents of S and P in clauses and so on. Moreover,
in each language, the system is one of two terms — singular and
plural — and these terms may also be regarded as formally cor-
responding. The exponents of the terms are differently distributed
in the two languages — e.g. Eng. the case /the cases — Fr. le castles cas
— but as terms in a number system singular and plural correspond
formally at least to the extent that in both languages it is the
term plural which is generally regarded as morphologically
marked.
In translation, however, it quite frequently happens that this
formal correspondence is departed from, i.e. where the trans-
lation equivalent of English singular is French plural and vice-
versa.
e.g.
advice =
news =
lightning =
applause =
trousers =
the dishes =
the contents =
des conseils
des nouvelles
des Eclairs
des applaudissements
le pantalon
la vaisselle
le contenu etc."
Again, we might regard English and French as having formally
• cf. Vinay et Darbelnet, pp. 119-23.
80
TRANSLATION SHIFTS
corresponding systems of deictics, particularly articles-, each may
be said to have four articles, zero, definite, indefinite and partitive.
It is tempting, then, to set up a formal correspondence between
the terms of the systems as in this table:
French
Zero —
Definite le, la, 1’, les
Indefinite un, une
Partitive du, de la, de I’, des
English
the
a, an
some, any
In translation, however, it sometimes happens that the equivalent
of an article is not the formally corresponding term in the system:
eg-
II est — professeur.
II a la jambe cassee.
Z/amour
Du vin
He is a teacher.
He has a broken leg.
Love
Wine
In the following table we give the translation-equivalents of
French articles found in French texts with English translations.
The number of cases in which a French article has an English
equivalent at word-rank is 6958, and the figures given here are
percentages; the figure 64 6 against le for instance, means that
the French definite article (le, la, 1’, les) has the English definite
article as its translation equivalent in 64-6% of its occurrences 7 .
By dividing each percentage by 100 we have equivalence-
probabilities — thus we may say that, within the limitations stated
above, French le, etc., will have Eng. the as its translation
equivalent with probability -65.
French English
zero
the
some
a
(other)
zero
67'7
6-1
0-3
11-2
4-6
le
14-2
64 6
—
2-4
18-9
du
513
9-5
110
5-9
22-4
un
6-7
5-8
2-2
70-2
15-1
7 1 am indebted to Dr. R. Huddleston for this information.
81
A LINGUISTIC THEORY OF TRANSLATION
It is clear from this table that translation equivalence does not
entirely match formal correspondence. The most striking diver-
gence is in the case of the French partitive article, du, the most
frequent equivalent of which is zero and not some. This casts doubt
on the advisability of setting up any formal correspondence
between the particular terms of the English and French article-
systems.
82
13
Language Varieties in Translation
The concept of a ‘whole language’ is so vast and heterogeneous
that it is not operationally useful for many linguistic purposes,
descriptive, comparative and pedagogical. It is, therefore, desir-
able to have a framework of categories for the classification of
‘sub-languages’, or varieties within a total language; that is,
idiolects, dialects, registers, styles and modes.
13.1 In theory, a ‘whole language’ may be described in terms
of a vast inventory of grammatical, lexical, phonological and, in
some cases, graphological forms, together with information about
relevant substance (e.g. features of phonic substance utilized in
the phonology), and statistical information (on relative frequency
of forms, etc.). 1 All of these may be said to constitute features of
the language.
Within this theoretical total inventory of features we can
establish sub-groupings or sub-sets of features. Such sub-sets might
be made up more or less at random — for example a sub-set of
English items like the following:
arthropoda, ashet, bedight, caitiff, cannot, can't, outwith, triploblaslic.
All the items in this sub-set are ‘English’ in the sense that all
may be found in English texts. On the other hand, it is difficult
to see much value in a grouping such as this — except for the
specific purpose of illustration, as here, or perhaps as an exam-
ination-item (as a test of the candidate’s ability to recognize
varieties of English) .
13.11 For most linguistic purposes it is desirable to establish
sub-sets of ‘features’ characteristic of varieties of the language
which regularly correlate with certain broad contextual or
situational categories. It is clear that the items listed above can
be grouped in such a way. Arthropoda and triploblaslic are charac-
1 Whether this ‘inventory’ is, indeed, an inventory or systematic listing (as
in ‘taxonomic’ description) or an ordered set of rules (as in ‘transformational’
description) is irrelevant. In either case the description is unmanageably vast.
83
A LINGUISTIC THEORY OF TRANSLATION
teristic of scientific, specifically zoological, English. Ashet and
outwith are characteristically Scottish — they occur in texts written
or spoken by Scotsmen. Bedight and caitiff are archaic — they are
most characteristically found in texts written a few centuries ago.
For many users of English cannot is characteristically a written
rather than a spoken form. The form can't may also occur in
written texts, but it differs from cannot in that it correlates with
situations in which there is a greater degree of familiarity between
the writer and his reader (s).
13.2 A language variety, then, is a sub-set of formal and/or
substantial features which correlates with a particular type of
socio-situational feature. For a general classification of varieties
we confine ourselves to a consideration of situational correlates
which are constants in language-situations. These constants are
(i) the performer (speaker or writer), (ii) the addressee (hearer or
reader) , and (iii) the medium (phonology or graphology) in which
the text is presented.
These three are ‘constants’ in the sense that they are invariably
present, or implied, in all language-situations. Performer and
addressee are socio-linguistic roles — whether or not both roles are
played by different individuals is quite irrelevant. A man may
talk to himself, in which case he is simultaneously filling the roles
of performer and addressee; or a broadcaster may talk into a
‘dead’ microphone, unaware that a breakdown has occurred,
in which case there are no listeners to fill the role of addressee,
but the addressee-role is still implicit in the performer’s selection
of language-material. Finally, every text is externalized in some
form or other— the performer must always select one or other
medium in which to expound the grammatical/lexical forms he
is using.
Varieties fall into two major classes: (i) those which are more
or less permanent for a given performer or group of performers, and
(ii) those which are more or less transient in that they change with
changes in the immediate situation of utterance. The major
varieties are listed in 13.21, and discussed in more detail in
13.4 onwards.
13.21 Types of variety related to permanent characteristics of
the performer(s).
84
LANGUAGE VARIETIES IN TRANSLATION
13.211 Idiolect: language variety related to the personal iden-
tity of the performer.
13.212 Dialect: language variety related to the performer’s
provenance or affiliations in a geographical, temporal or social
dimension.
(i) Dialect ( proper ) or Geographical Dialect: variety related to the
geographical provenance of the performer: e.g. ‘American
English’, ‘British English’, ‘Scottish English’, ‘Scots Dialect’.
(ii) Dtat de langue or Temporal Dialect: variety related to the
provenance of the performer, or of the text he has produced, in
the time dimension: e.g. ‘Contemporary English’, ‘Elizabethan
English’, ‘Middle English’.
(iii) Social Dialect: variety related to the social class or status of
the performer: e.g. ‘U and non-U’ (U = Upper Class).
13.22 Types of variety related to ‘transient’ characteristics of
the performer and addressee — i.e. related to the immediate
situation of utterance.
13.221 Register: variety related to the wider social role being
played by the performer at the moment of utterance: e.g.
‘scientific’, ‘religious’, ‘civil-service’, etc.
13.222 Style: variety related to the number and nature of
addressees and the performer’s relation to them: e.g. ‘formal’,
‘colloquial’, ‘intimate’.
13.223 Mode: variety related to the medium in which the
performer is operating: ‘spoken’, ‘written’.
13.23 It would, no doubt, be possible to add to this list of
variety-types, particularly by sub-division or conflation. For
instance, a more delicate classification of medium-manifestation
might supply ‘secondary modes’ — such as ‘telegraphese’, a sub-
type of the written mode. Again, a kind of conflation might
provide us with a ‘poetic genre’ as a super-variety characterized
by potential use of features appropriate to all varieties. For the
present study, however, we confine ourselves to the varieties
listed here.
13.3 All languages may be presumed to be describable in terms
of a number of varieties, though the number and nature of these
varies from one language to another — a fact of importance in
connection with translation.
85
A LINGUISTIC THEORY OF TRANSLATION
All the varieties of a language have features in common — these
constitute a common core of e.g. grammatical, lexical and phono-
logical forms. In addition to the common core, however, every
variety has features which are peculiar to it, and which serve as
formal (and sometimes substantial) criteria or markers of the
variety in question.
The markers of particular varieties may be at any level:
phonetic, phonological, graphological, grammatical, lexical. As
far as dialect is concerned, many languages have a ‘standard’ or
‘literary’ dialect, which shows little variation (in its written form
at least) from one locality to another. It is convenient, particularly
in connection with translation, to regard such a dialect as
unmarked.
13.4 An idiolect is the language variety used by a particular
individual. The markers of an idiolect may include idiosyncratic
statistical features, such as a tendency to exceptionally frequent
use of particular lexical items. A person’s idiolect may change in
detail from day to day (e.g. by the adoption of ‘new pronuncia-
tions’, the acquisition of new lexical items, etc.), and may change
extensively in a life-time. For most purposes, however, we may
regard an adult’s idiolect as relatively static.
It is not always necessary to attempt to translate idiolects:
i.e. the personal identity of the performer is not always an
important feature of the situation. It may happen, however, that
the performer’s identity is relevant. For instance, in a novel,
idiolectal features in the dialogue of one character may be worked
into the plot; other characters may remark on these, and they
may partly serve to identify the character. In such a case the
translator may provide the same character in his translation with
an ‘equivalent’ idiolectal feature 2 .
13.5 A dialect , as we have seen, is a language variety, marked
by formal and/or substantial features relatable to the provenance
of a performer or group of performers in one of the three dimen-
sions — space, time and social class.
» Those features of what is often called the individual ‘style’ of a particular
author are idiolectal, and in a literary translation some attempt may have to
be made to find TL equivalents for them. Unusual collocations may also
sometimes be regarded as idiolectal — for an example see 14.52.
86
LANGUAGE VARIETIES IN TRANSLATION
13.51 Geographical dialects may be defined with greater or lesser
specificity ; thus among dialects of English, for instance, we may
for some purposes distinguish between British English, American
English, Australian English, etc. — such broad, inclusive, dialects
being formally distinguished from one another by relatively few
markers. For other purposes we may specify sub-varieties within
these broad categories, e.g. Scots English, and within this, again,
still more strictly localized varieties. Similarly, e'tats de longue may
be arbitrarily marked off along the time-dimension very broadly,
as Old English, Middle English, Modern English, etc., or more
strictly located within these broad periods, e.g. 19th Century
English . . .
13.52 Dialects may present translation problems. As we have
said in 13.3 above for most major languages there is a ‘standard’
or ‘literary’ dialect which may be regarded as unmarked. Texts
in the unmarked dialect of the SL can usually be translated in an
equivalent unmarked TL dialect. When the TL has no equivalent
unmarked dialect the translator may have to select one particular
TL dialect, create a new ‘literary’ dialect of the TL, or resort to
other expedients. This problem not infrequently arises in the
translation of the Bible into pre-literate languages, and has been
discussed at some length by E. A. Nida 3 .
13.53 When an SL text contains passages in a dialect other
than the unmarked dialect (e.g. in the dialogue of novels) the
translator may have to select an equivalent TL dialect. Translation
equivalence, as we have seen in 7.4 above, depends on relation-
ship of SL and TL text to ‘the same’ substance; for total trans-
lation, this is situation substance. In the selection of an equivalent
TL geographical dialect this means selection of a dialect related
to ‘the same part of the country’ in a geographical sense. Geo-
graphy is concerned with more than topography and spatial
co-ordinates — and human geography is more relevant here than
mere location. Thus, in relation to the dialects of Britain, Cockney
is a south-eastern dialect. In translating Cockney dialogue into
French, however, most translators would select Parigot as the TL
equivalent dialect, even though this is a northerly dialect of
* Bible Translating, Chapter 3.
87
A LINGUISTIC THEORY OF TRANSLATION
French. The criterion here is the ‘human’ or ‘social’ geographical
one of ‘dialect of the metropolis’ rather than a purely locational
criterion.
13.54 The markers of the SL dialect may be formally quite
different from those of the equivalent TL dialect. There are
certainly Cockney markers at all levels, but in many literary
texts, Cockney is marked chiefly by phonological features,
reflected in such graphological forms as ’alf or ’arf for ‘half’, wiv
for ‘with’, and a few grammatical features such as aint for ‘isn’t/
aren’t’. In addition there are often pseudo-phonological features,
indicated by graphological peculiarities such as orful for ‘awful’
and ter for ‘to’. These graphological forms can be interpreted
phonologically only in perfectly normal ‘standard’ English ways
— they are the markers of a purely visual, literary, dialect.
The markers of Parigot may include a few graphological/
phonological features, but are likely to be largely at the lexical
level, in the form of extensive use of argot. This illustrates another
case, like those cited in 7.6 above, where phonological/grapho-
logical features appear to have translation equivalents at the level
of lexis; but, as in those examples, this is an illusion. If the
translation equivalent of Vs gorn is il a foutu I'camp this does not
mean that lexical items are here translation equivalents of
phonological features. The translation equivalence is set up
between varieties (specifically Cockney and Parigot ) : of which the
SL markers are phonological, and the TL markers are lexical;
there is no equivalence between phonological and lexical features
as such.
13.55 Temporal dialects, or etats de langue, may also present
translation problems. A contemporary etat de langue of the SL may
normally be regarded as unmarked, only archaic varieties being
marked. In spoken-spoken translation 4 both SL and TL texts are
normally ‘contemporary’ or ‘unmarked’ dialects in the time-
dimension. An archaic SL text, however, raises the problem of
4 Here we imply a categorization of ‘external’ aspects of translation not dealt
with in the present essay. For this categorization at least four dimensions have
to be considered; viz. those of media (SL spoken/written, TL spoken/written),
time-relation (simultaneous/successive), agent (human/machine), and for human
translation at least, direction (L, to L,, or L, to LJ.
88
LANGUAGE VARIETIES IN TRANSLATION
whether, and how, the translator should seek to select an equi-
valent archaic TL text. Here, as in the case of geographical
dialect, equivalence of absolute location in time is normally
neither possible nor desirable. The 12th Century Russian Slovo
o polku Igoreve, for example, would not normally be translated into
12th Century English; in this form it would be considerably less
intelligible to a contemporary English reader than the original
is to a contemporary Russian. Dennis Ward 5 in his article on
translation of the Slovo has argued against archaism in the TL
text, with the exception of his deliberate selection of host as
translation equivalent of the lexical item polk. Nevertheless, parts
of his brilliant translation have a somewhat ‘archaic flavour' — -
the markers here being lexical items such as girded, beasts , want >rs,
behold, the use of brothers as a term of address, not to mention
lexical items which are of low frequency in contemporary English
texts because their contextual meanings relate to archaic objects
or institutions such as bows, quivers, shields, pennons, gerfalcon, pagan
hosts', occasional features of clause-structure; phonological fea-
tures of alliteration and metre (successful partial phonological
translations) and so on. Such features are for the English reader,
markers of a slightly archaic etat de langue, appropriate to the
subject as well as being to some extent a translation equivalent
of the SL etat de langue.
13.6 Register, Style and Mode are language varieties related to
the immediate situation of utterance.
13.61 By register we mean a variety correlated with the per-
former’s social role on a given occasion. Every normal adult plays
a series of different social roles — one man, for example, may
function at different times as head of a family, motorist, cricketer,
member of a religious group, professor of biochemistry and so on,
and within his idiolect he has varieties (shared by other persons
and other idiolects) appropriate to these roles. When the pro-
fessor’s wife tells him to ‘stop talking like a professor’ she is
protesting at a misuse of register.
13.611 Registers, like dialects, can be defined with lesser or
s ‘On translating Slovo o polku Igoreve’, The Slavonic Review. The translation
itself ‘The Tale of the Host of Igor . . .’ appeared as a supplement to Ward’s
paper to the IVth International Congress of Slavists, Moscow 1958.
89
A LINGUISTIC THEORY OF TRANSLATION
greater specificity; thus, while in English we may identify a
general scientific register, we may also differentiate sub-registers
within this. Register-marArerr are chiefly lexical (most obviously
‘technical terms’, but including other items), and grammatical,
particularly grammatical-statistical features such as the high
frequency (30% to 50%) of passive verb forms, and the low
frequency of the pronouns I you he and she in English scientific
register.
13.612 In translation, the selection of an appropriate register
in the TL is often important. Here, if the TL has no equivalent
register, untranslatability may result. One of the problems of
translating scientific texts into certain languages which have
recently become National Languages, such as Hindi, is that of
finding, or creating, an equivalent scientific register. And here
again, the equivalence is between varieties ; an English scientific text
may have, inter alia, a relatively high percentage occurrence of
passives-, its Russian translation a relatively high percentage
occurrence of javlaels'a -f- instrumental. The Russian javlaets'a is
not necessarily the translation equivalent of an English passive ;
both arc merely markers of equivalent registers.
13.62 By style we mean a variety which correlates with the
number and nature of the addressees and the performer’s rela-
tionship to them. Styles vary along a scale which may be roughly
characterized as formal . . . informal. For English, Martin Joos
has suggested five styles: frozen, formal, consultative, casual and
intimate?.
The markers of styles may be lexical, grammatical or phono-
logical. Not much is known in detail about English styles, though
it is probably true, as Joos points out, that ellipsis is one marker
of casual style: e.g. Coffee's cold. Bought it yesterday. Leaving ? —
another is the use of slang. For English we can probably regard
consultative style as the unmarked style in the spoken mode,
though formal style may be the unmarked style in the written
mode.
As with registers, so with styles, translatability depends on the
existence of an equivalent style in the TL. In English, style-
• ‘The Isolation of Styles’, Georgetown Monographs on Languages and Linguistics
12 (1959), pp. 107-13.
90
LANGUAGE VARIETIES IN TRANSLATION
markers tend to be dispersed over a number of levels of the
language, including lexis and phonology. In many languages,
particularly in South East Asia, the translation equivalents of
particular English styles may be more rigidly built into grammar
and lexis — as the use of specifically ‘self-abasing’ or ‘honorific’
terms in a system of pronouns, or similar obligatory alternative
items in lexical sets.
Here again, however, translation equivalence must be set up
between the varieties as such, and the specific markers may be
very different in the SL and TL texts. Moreover, the equivalence
is ultimately based on similarities of situation-substance — only,
those which are stylistically relevant in one language may not be
in another. An English youth may easily address his father in
casual style; an oriental youth on the other hand may have to
use honorific forms in such a situation. Both respect and affection
may be present in the situation, but respect may not be a
stylistically relevant feature for the English son, while it is rele-
vant for the Asian son.
This is one reason for divergences here, as elsewhere, between
formal-correspondence and translation equivalence. Two lan-
guages might possess a roughly corresponding set of styles; but
cultural factors may dictate the use of a non-corresponding style
as translation equivalent.
13.7 It should be noted that there may be syncretisms and
incompatibilities between varieties. For one thing, in English, as
we move ‘down’ the style-scale from formal to casual the registral
differences become less marked. A professor of zoology may give
a lecture to a learned society in zoological register and formal
style. He may continue to use zoological register with the con-
sultative style he uses in a seminar with graduate students, or
with the casual style he uses in common-room scientific gossip
with colleagues. Specific lexical items — the ‘technical terms’ or
zoology-will still be there as register-markers in his casual style,
but most of the other markers of scientific register— the less
specifically zoological, but still scientific, lexical items, the
grammatical markers and so on — will have disappeared.
There may be incompatibility between, say, a rural dialect and
scientific register, or between casual style and religious register
91
A LINGUISTIC THEORY OF TRANSLATION
and so on. Such incompatibilities may have an effect on trans-
lation. Thus some Hindi translations of English novels and short
stories show no attempt to use a particular Hindi ‘marked
dialect’ as translation equivalent of rural (geographical) dialect
or ‘uneducated’ social dialect in English dialogue. It is possible
that this reflects a dialect/mode incompatibility in Hindi — i.e.
the non-compatibility of ‘sub-standard’ Hindi dialect with the
written mode.
In many cases a change of style or register involves a corres-
ponding change of dialect or even language. In Arabic, for
example, the Classical dialect is hardly compatible with casual
style. Many Indians will switch from, say, Hindi or Marathi to
English whenever they speak or write about scientific subjects;
such people have no scientific register in their ‘mother tongue’,
but only in English.
92
14
The Limits of Translatability
14.1 In 7.6 above we were able to state certain absolute limits
of translatability, namely: translation between media is impos-
sible, and translation between the medium-levels and the levels
of grammar/lexis is likewise impossible. These absolute limitations
derive directly from our theory of translation equivalence. For
translation equivalence to occur, SL and TL items must be
relatable to (at least some of) the same features of substance, and
it is easy to see that there is an absolute absence of similarity
between phonic and graphic substance, and between either of
these and situation substance.
14.11 The limits of translatability in total translation are,
however, much more difficult to state. Indeed, translatability
here appears, intuitively, to be a dine rather than a clear-cut
dichotomy. SL texts and items are more or less translatable rather
than absolutely translatable or untranslatable. In total translation,
translation equivalence depends on the interchangeability of the
SL and TL text in the same situation — ultimately, that is, on
relationship of SL and TL texts to (at least some of) the same
relevant features of situation-substance.
14.12 At this point we must consider more closely the term
relevant. In Chapter 7 we talked about linguistically relevant features
of situation (substance) — those features, or bundles of features,
which led to the performer selecting this or that item of his
language. Similarly, in the example in 5.4 we saw that for a
Russian speaker, the sex of the performer was linguistically
relevant, that is, led to selection of the form prisla as opposed to
prisel. For the equivalent English text the sex of the performer
was linguistically irrelevant — i.e. did not lead to selection of one
particular linguistic form rather than another.
The English and Russian texts, I've arrived and ja prisla operate
perfectly well as translation equivalents in spite of this difference,
because the sex of the performer though linguistically relevant
for the Russian text is not relevant to the communicative function
93
A LINGUISTIC THEORY OF TRANSLATION
of the text in that situation ; in other words, the Russian performer
is obliged by a formal feature of her language to make this
incidental reference to her sex, even though this is not ‘what she
intends to say’.
14.13 We can distinguish, then, between situational features
which are linguistically relevant, and those which are functionally
relevant in that they are relevant to the communicative function
of the text in that situation. For translation equivalence to occur,
then, both SL and TL text must be relatable to the functionally
relevant features of the situation. A decision, in any particular
case, as to what is functionally relevant in this sense must in our
present state of knowledge remain to some extent a matter of
opinion. The total co-text will supply information which the
translator will use in coming to a decision, but it is difficult to
define functional relevance in general terms.
14.14 Translation fails — or untranslatability occurs — when it
is impossible to build functionally relevant features of the situa-
tion into the contextual meaning of the TL text. Broadly
speaking, the cases where this happens fall into two categories.
Those where the difficulty is linguistic, and those where it is
cultural.
14.2 In linguistic untranslatability the functionally relevant fea-
tures include some which are in fact formal features of the
language of the SL text. If the TL has no formally corresponding
feature, the text, or the item, is (relatively) untranslatable.
Linguistic untranslatability occurs typically in cases where an
ambiguity peculiar to the SL text is a functionally relevant feature
— e.g. in SL puns.
14.21 Ambiguities arise from two main sources, (i) shared
exponence of two or more SL grammatical or lexical items, (ii)
polysemy of an SL item with no corresponding TL polysemy.
14.211 By shared exponence we mean those cases where two or
more distinct grammatical or lexical items are expounded in one
and the same phonological or graphological form.
A grammatical example in English is the shared exponence of
the two distinct morphemes ‘(Nominal) plural’ and ‘(Verbal)
third person singular present’ both of which are frequently
expounded graphologically by -s, as in cats and eats. In most cases,
94
THE LIMITS OF TRANSLATABILITY
there is no ambiguity, since the co-text (as here) indicates clearly
which item is being expounded, and the translation equivalent
is then not in doubt. But cases of ambiguity can arise, an example
is Time flies. If this piece of text occurred in a normal conversation
there would be no translation problem; the co-text would show
whether the contextual meaning was ‘How quickly time passes’,
or something like ‘Make observations on the speed of flies’, and
the appropriate translation equivalent would be obvious. But
when the whole point of the text is to provide an example of
ambiguity, as it is in this paragraph, then translation is virtually
impossible. The ambiguity itself (a feature of the English language
—the SL) is a functionally relevant feature of the situation.
A lexical example might be bank, which is the graphological
exponent of two distinct lexical items in English. 1 This normally
presents no problem in translation; the co-text normally shows
whether, for example, the French translation equivalent should
be banque or rive. But bank is untranslatable when the ambiguity
is itself a functionally relevant feature, as in Ogden and Richard’s
punning fable about Amoeba 2 , which begins:
‘Realize thyself, Amoeba dear’, said Will ; and Amoeba realized
herself, and there was no Small Change but many Checks on the
Bank wherein the wild Time grew and grew and grew.’
Here it is clear that the reader is expected to relate the grapho-
logical form Bank to both the lexical items which it expounds.
This is impossible in French, where the translation equivalent
must be either banque or rive and not both at once; and other
untranslatable ambiguities are equally obvious in this text.
14.212 The second type of linguistic ambiguity is due to what
would usually be called polysemy, that is, not to the fact that two
or more items have the same exponent, but that one single item
has more than one meaning. Strictly speaking, the term polysemy
1 That bank represents two items — not just ‘one item with two meanings’ —
is intuitively felt by English speakers. The formal confirmation of this intuition
will no doubt be forthcoming when computers have demonstrated that ‘bank’
occurs in two slightly overlapping but largely quite distinct collocational
ranges.
2 C. K. Ogden and I. A. Richards, The Meaning oj Meaning, Appendix E.
95
A LINGUISTIC THEORY OF TRANSLATION
is misleading. It is not a case of one item having several meanings,
but of one item having a wide or general contextual meaning,
covering a wide range of specific situational features. In any given
situation, only one out of this wide range of potentially, or
linguistically, relevant features is functionally relevant. An
example is the Russian s verxu the contextual meaning of which
can be roughly summarized in English as being ‘from or off a
higher position’. S verxu is thus appropriate to situations in which
the English translation equivalent would be ‘from above’, ‘from
upstairs’, ‘from upriver’ . . . etc. Normally, the co-text shows
which part of the total contextual meaning of s verxu is function-
ally relevant, and translation presents no problem. But on rare
occasions the linguistic feature itself, the wideness of meaning of
the item s verxu — its polysemy — is a functionally relevant feature.
In this case, translation is virtually impossible — an example is
given in 14.31 below.
14.22 In addition to ambiguity, due to shared exponence or to
polysemy, another kind of linguistic untranslatability can occur.
In this case it is not polysemy, but rather what might be called
oligosemy which is the cause.
If an SL item has a particularly restricted range of meaning
it may not be possible to match this restriction in the TL.
Normally, again, this does not matter. The Russian prilla, as we
saw above, means ‘came’ or ‘arrived’ on foot. English has no
lexical item with a correspondingly restricted range of contextual
meaning; but this does not prevent English came or arrived from
often being a perfect translation equivalent. In special cases,
however, this restriction of meaning — the ‘oligosemy’ of prisla as
opposed to English came — may itself be a functionally relevant
feature of the situation. This, like the previous type of ambiguity,
is illustrated in 14.3 below.
14.3 Examples of (relative) linguistic untranslatability due to
all the factors dealt with above are well illustrated in the following
passage in Maxim Gorki’s Childhood.
14.31 The child, Gorki, has been ill in bed for some time. His
grandmother has travelled down the Volga from Nijni Novgorod
to look after the family. To the little boy, she is just a new grown-
up who has suddenly appeared on the scene. The following
96
THE LIMITS OF TRANSLAT ABILITY
conversation occurs, presented here with a rank-bound (largely
word-word, partly morpheme-morpheme) and unbounded
translation.
Ty otkuda prisla.
Thou whence came-on-foot?
Where have you come from ?
S verxu, iz Niznego, da ne prisla,
From above, from Lower and not came-on-foot.
From upriver/upstairs, from Nijni/lower, and I didn’t come
on foot.
Po vode-to ne xod'at.
Onwater-(!) not they-go-on-foot.
You don’t walk on water!
The child finds this funny and confusing; he reflects on who
lives upstairs and downstairs in the house, and he wonders how
one can come down the stairs without going on foot; and what
has water to do with it ?
14.32 The untranslatability of this text, or rather of certain
items in it, has nothing to do with cultural differences in the
wider sense; it is purely linguistic. It rests on the SL items prisla ,
s verxu , Niznego, and these illustrate all three of the causes of
linguistic untranslatability referred to above.
14.321 Niznego illustrates ambiguity arising from shared ex-
ponence — niznij (genitive singular niznego ) as exponent of (a) an
adjective meaning lower and (6) the common abbreviation of the
place-name Nijni Novgorod (‘Lower Newtown’). This is virtually
untranslatable into English because a comparable shared ex-
ponence does not occur — Lower may occur as a place-name
element, but it is not normally used by itself as an abbreviation.
S verxu is an example of ambiguity due to ‘polysemy’ or the wide
range of contextual meaning of this item. Out of the total range
of situational features with the general characteristic of being
‘from above’, the child selects the specific feature ‘from above in
the house’, or ‘from upstairs’, while the grandmother means
‘from up yonder’, or ‘from upriver’. English cannot easily com-
bine these specific features in the contextual meaning of one
lexical item; it must select ‘from upstairs’, or ‘from up yonder’,
97
A LINGUISTIC THEORY OF TRANSLATION
or ‘from upriver’. The equivalent ‘from above’ would be colloca-
tionally strange in this text. Finally, prilla illustrates a case of
‘oligosemy’. The item prisla means ‘came’ or ‘have come’ on foot.
In many situations, the situational feature referred to by ‘on foot’
is not functionally relevant, though it is linguistically relevant for
Russian. Consequently, a perfectly good English translation can
often ignore this feature and use the English come which has a
wider situational range. In this example, however, this feature,
which is linguistically relevant for Russian, is also functionally
relevant, since it is an important factor in causing the child’s
bewilderment.
14.322 We might attempt a more ‘faithful’ translation some-
what on these lines:
Where have you walked in from ?
I’ve just come down — from Lower.
And I didn’t walk. You don’t walk on water.
It is clear, however, that this translation is unsatisfactory. The
sentence ‘Where have you walked in from?’ is out of register.
‘From Lower’ would convey nothing to an English reader without
a footnote explaining that ‘Lower’ is a translation of the abbre-
viated form of Nijni Novgorod. And, finally, ‘I’ve just come down’
(or any obvious alternative) does not suggest the quite specific
interpretation ‘from upstairs’ which the child Gorki gives to the
Russian s verxu.
14.4 The ‘untranslatability’ exemplified in the last few para-
graphs is called linguistic untranslatability because failure to
find a TL equivalent is due entirely to differences between the
source language and the target language. Such differences are, of
course, the rule rather than the exception, since formal corres-
pondence is exceedingly rare— but formal differences between
languages do not normally preclude the finding of translation
equivalents. Formal linguistic differences— differences between
the SL and the TL organization of situation-substance — lead to
translation failure only when the SL formal feature is itself a
textually-functionally-relevant feature. The related situational
features may themselves be perfectly commonplace in both the
SL and TL cultures.
98
THE LIMITS OF TRANSLATABILITY
What appears to be a quite different problem arises, however,
when a situational feature, functionally relevant for the SL text,
is completely absent from the culture of which the TL is a part.
This may lead to what we have called cultural untranslatability.
This type of untranslatability is usually less ‘absolute’ than
linguistic untranslatability.
14.41 We have already referred in passing to a Finnish lexical
item which may be untranslatable into English — namely sauna
(see 6.31 above). There may be texts in which bath or bathhouse
would be an adequate translation equivalent. But the Finnish
and the English institutions are certainly different, and a sauna
is not always a separate building — it may be a room in a house,
hotel, or ship for instance. In this latter case, the obvious English
equivalent bathroom would probably be evaluated by any trans-
lator as inappropriate.
It is a curious fact that the Japanese lexical item huro(-ba) seems
to be more easily translatable as bath or bathroom than the Finnish
sauna. And yet the Japanese bath(room) is in some respects as
different from an English bath(room) as is the sauna — and both
of the non-English institutions have non-English features in
common.
As distinct from the English bath, which is normally a solitary
activity, the Finnish and Japanese baths are, or may often be,
communal. The Finnish and Japanese ‘bathrooms’ are, each in
its own way, quite differently constructed and furnished from an
English bathroom.
The sauna, however, differs still more (has more non-English
situational features) from the English bath or bathroom; it
involves neither immersion in hot water, nor washing the body
(which is done outside the sauna and is not an integral part of
‘taking a sauna’). The Japanese institution, like the English one,
does involve immersion in hot water, and washing the body is an
integral part of the bath-taking and is performed inside the
bathroom itself, though before actually entering the water to soak.
It looks, therefore, as if equivalence of material aspects of the
institution are less important than equivalence in its major
personal or social function (washing the body and soaking in hot
water) in promoting translatability. This reminds us of the point
99
A LINGUISTIC THEORY OF TRANSLATION
referred to in 13.53 above where it was suggested that the human
or socio-geographical status of a SL dialect might be the essential
situational feature determining the selection of an equivalent TL
dialect, rather than its geographical location.
14.42 Articles of clothing provide other examples of features of
material culture which differ from one culture to another and
may lead to translation difficulties. The contextual meaning of
the Japanese lexical item yukata, for example, includes some such
features as ‘loose robe bound by a sash, worn by either men or
women, supplied to guests in a Japanese inn or hotel, worn in the
evening indoors or out of doors in street or cafe, worn in bed . . .’
etc. Parts of this total range are covered by such English lexical
items as dressing-gown, bath-robe, house-coat, pyjamas, night-gown . . .
etc., and in some texts the relevant situational features might be
just those common to both dressing-gown and yukata on that
particular occasion. But no English item is relatable to the full
range of situational features, and there are likely to be texts where
no possible English translation equivalent exists. No English
garment, for instance, is worn both in bed and in the street
(except in emergencies) and certainly no garment is supplied by
English hotels to their guests.
The solution adopted by most translators here would be to
transfer the SL item yukata into the TL text, leaving its contextual
meaning to emerge from the co-text (or else explaining it in a
footnote). Another possibility would be to use the item kimono as
translation equivalent, since this originally Japanese lexical item
is already ‘naturalized’ as a loan-word in English, though yukata
and kimono do not mean the same in Japanese.
14.43 It is often supposed that certain more ‘abstract’ lexical
items such as home or democracy are relatively untranslatable. This
is largely an illusion. There is no doubt that such English texts
as He’s at home or I’m going home can readily be provided with
translation equivalents in most languages. It is only rarely that
the functionally relevant situational features related to home
include that nebulous sentimentality which is supposed not to be
related to lexical items in other languages — e.g. perhaps in the
song Home. Sweet Home.
As for democracy, this is in any case an international term —
100
THE LIMITS OF TRANSLATABILITY
which means essentially that it is untranslatable because it often
need not be translated — since it is already present in the lexis of
many languages ; an ‘international term’ being a lexical item with
recognizably similar phonological /graphological exponents in
several languages, and having a common contextual meaning.
The total range of situational features relatable to the contextual
meaning of democracy includes features which are present in some
national and political situations but absent from others — the co-
text generally guides the reader to selection of the appropriate
situational features in any particular case. Even within one
and the same language, democracymay be relatable to some different
situational features in the registers of different political parties.
14.5 Although we have, following a somewhat obvious and
intuitive approach, distinguished between linguistic and cultural
untranslatability it may be questioned whether such a distinction
is ultimately necessary. In many cases, at least, what renders
‘culturally untranslatable’ items ‘untranslatable’ is the fact that
the use in the TL text of any approximate translation equivalent
produces an unusual collocation in the TL. To talk of ‘cultural
untranslatability’ may be just another way of talking about
collocational untranslatability: the impossibility of finding an
equivalent collocation in the TL. And this would be a type of
linguistic untranslatability.
We might define collocational untranslatability thus: untrans-
latability arising from the fact that any possible TL near-equiva-
lent of a given SL lexical item has a low probability of collocation
with TL equivalents of items in the SL text which collocate
normally with the given SL item.
14.51 Thus, in the Japanese text hoteru-no yukata, the item
holeru-no has the straightforward English translation equivalent
hotel’s or hotel (as modifier); but any possible English near-
equivalent of yukata would collocate strangely with hotel — i.e.
hotel dressing-gown, hotel bath-robe, hotel nightgown, etc., are all low
probability collocations in English — though the original Japanese
collocation is a normal, or high-probability one.
More extended examples will make this point even clearer. The
following two texts are imaginary translations from Finnish and
Japanese respectively.
101
A LINGUISTIC THEORY OF TRANSLATION
(i) ‘They lay on the hot upper benches of the bathroom inhaling
the aromatic scent of the birch twigs.’
(ii) ‘After his bath he enveloped his still-glowing body in the
simple hotel bath-robe and went out to join his friends in the cafe
down the street.’
14.511 Both of these would ‘read strangely’ to an English
reader unacquainted with Finnish or Japanese institutions. This
strangeness can be attributed to the strangeness of the situations
they suggest — to the mild ‘cultural shock’ induced by the image
of (i) people (more than one) lying about on hot benches in a
birch-scented bathroom and (ii) of a hotel bath-robe which,
moreover, is worn in the street. We can, in other words, say that
bathroom and bath-robe are bad translations, and if no other English
lexical items, less suprising in these co-texts, can be found — then
we may say that the SL items sauna and yukata are untranslatable
— for cultural reasons.
14.512 But we can also describe the strange effect produced by
these translations not as ‘cultural shock’ but as ‘collocational
shock’. In other words we can attribute the relative untranslata-
bility of the two SL items to a purely formal linguistic feature —
unusualness of collocation. In theory, this could be done without
any appeal at all to the contextual meanings of the texts — and
hence without any reference to cultural differences. If a sufficient
amount of information were available on the collocation of
lexical items in any pair of SL and TL languages the ability to
identify such so-called ‘culturally untranslatable’ items might, in
theory, be programmed into a computer for the purposes of
machine translation.
14.52 The case is different with the following (genuine) trans-
lation from French. The SL text is a sentence from ‘La Chatte’,
by Colette®. The English TL text reads :
‘ The sun kindles a crackling of birds in the gardens. ’
There are certainly strange, or low-probability, collocations
here. But in this case the strangeness of the collocations is not due
to ‘untranslatability’ — on the contrary, it is, indeed, an indication
* This example is taken from the essay on translation in J. G. Weightman,
On Language and Writing (Sylvan Press, 1947).
102
THE LIMITS OF TRANSLAT ABILITY
of a ‘good’ translation, because a very similar strangeness of
collocations exists in the original :
‘Le soleil allume un crepitement d’oiseaux dans lesjardins.’
In other words, the collocation soleil — allume — crepitement —
oiseaux is about as unusual as the collocation: sun — kindles —
crackling — birds. From this we may deduce that collocational
abnormality in the TL text is a symptom of (so-called ‘cultural’)
untranslatability only when the original SL text is collocationally
normal. When the SL text is itself collocationally abnormal an
equivalent collocational abnormality in the TL text may be
merely the mark of a ‘good’ translation.
In this particular example from Colette there is, as Weightman
points out, some degree of untranslatability. The French item
crepitement has certain associations for a French reader which are
— perhaps inevitably — lost in the English translation. The major
untranslatable ‘association’ of crepitement is that it is somewhat
reminiscent of pepiement, a lexical item used to refer to the twitter-
ing of birds. Now this untranslatable association of crepitement is a
good example of one of the types of linguistic untranslatability
referred to in 14.211 above, namely shared experience. The phono-
logical forms represented graphologically by crepitement and
pepiement are partially alike— in other words, we have here two
French lexical items with (partially) shared exponence. Whether
or not we regard the resultant simultaneous reference to situa-
tional features of the contextual meanings of both these items as
functionally relevant or not may be a matter of opinion. But if
we do accept this view, and if we do in consequence say that
crepitement is to some degree untranslatable, then we must accept
the fact that this is a case of linguistic untranslatability.
14.6 Here we have been able only to touch on the problem of
the limits of translatability. The subject is a large one and requires
much further study. If, indeed, it should turn out that ‘cultural
untranslatability’ is ultimately describable in all cases as a variety
of linguistic untranslatability, then the power of translation-theory
will have been considerably increased and, among other things,
the horizon of machine translation will have been enlarged.
103
A series bringing together writings
from the different fields of linguistics,
language study, and language teaching
methodology and materials.
*
J. C. Catford A Linguistic Theory of Translation
This is an important work which brings a new degree
of precision into the analysis of what is involved in
translation from one language to another. Starting from
v the assumption that any process concerned with
human language can be illuminated by applying to it
the latest insights into the nature of language, the
author outlines a current British frame-work of
descriptive linguistics and applies it to the analysis of
translation. Translation is shown to be a much more
complex matter than is commonly realized, while at
the same time the author indicates important new ways
of approaching it. The book is a valuable addition
to the literature of a subject which has only recently
begun to receive the scientific treatment it deserves.
LANGUAGE
Q>
ZJ
Q_
LANGUAGE
LEARNING
Oxford University Press
ISBN 0 19 437018 6
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Open Library
American Libraries
TV News
Understanding 9/11